FOUNDATIONS OF ELECTROMAGNETIC THEORY by JOHN R. REITZ Department of Physics Case Institute of Technology and FREDERICK J. MELFORD Mathematical Physics Division Battelle Memorial Institute ADDISON-WESLEY PUBLISHING COMPANY, ic. READING, MASSACHUSETTS, U.S.A. LONDON, ENGLANDCopyright © 1986 ADDISON-WESLEY PUBLISHING COMPANY, Ic. Printed in the United States of America ALL RIGHTS RESERVED. THIS HOOK, OR PARTS THERK- OF, MAY NOT BE REPRODUCED IN ANY TORM WITT- OUT WRITTEN PERMISSION OF THE PUBLISHER. Library. of Congress Catalop Curd No, 60-8247PREFACE Although Maxwell’s equations were formulated more than seventy years ago, the subject of Electricity and Magnetism has wot remained static. The progress of the 1930’e on the microscopic constitution of matter, and the growth of solid-state physics after World War II, have led to a better understanding of electric and magnetic fields inside matter. The alvanced undergraduate siudent in Science (thia is the student to whom we are directing our attention) upproaches the subject of electricity with a qualitative understanding of atomic phenomena. - At the same time, he has acquired a good background in mathematics, and for the first time in his career he is in the position of being able tc solve some of the im- portact mathematical problems of classical physics. “It appeared to us that there was no well-designed text in Electricity and Magnetism to meet the special needs of this group of students. ‘The present, vohume haa evolved froth the teaching of a course in Elec- tricity and Magnetism to physics majors at Case Institute of Technology. ‘These utudents have been introduced to vector analysis both in mathe- matics and mechanics courses, they have encountered some of the im- portant partial differential equations in physics, and -they have been invreduced to boundary-value problems. A course in Electricity and Magnetism is ideally suited to a ‘urvher development of these mathe- matical concepts, and we have attempted to exploit this ides in the present book. Although a previous introduction to these concepis is desirable, the sections on vector analysis and boundary-value problems have been wri in such a way thai little previous knowledge of the subject is required. We feel that the approach of building up Electricity and Magnetism from the basic experimental laws is still the correct one ut the intermediate level, and we have followed this approach. Although a rigorous exposition of the fundamentals is to be preferred to teaching by example, we have been careful to include a substantial number of well-chosen example problems to bridge the gap between the formul development of the subject and the problems. Mxperience has shown that « deficiency of examples can detract, from an otherwise good textbook. It is our belief thut a full understanding of the electric and magnetic fields inside matter cun be obtained ouly after the atomic nature of matter is appreciated. Hence we have used clementary atornic concepis freely in the development of macroscopic them'y, We ltive tried to use the physical approach in our treatment of polaridation aid fiagnetization, as well as in our discussion of Lhe auxiliary vectors D and'H. We believe our book has something extra to offer in this ared?,vi PREFACE Separate short chapters have been written on the microscopic theory of dielectric and magnetic material. This is a subject usually neglected in a book on formal theory, but it seems to us that many of the concepts involved are simple, and best presented early. The special features of our book are: (1) a full vector treatment of the subject, including the use of vector identities to simplify proofs of theo- rems, (2) a utilitarian approach to boundary-value problems and their solution, (3) a rigorous development of Electricity and Magnetisth from experimental laws, with no essential proofs relegated to more advanced texts, (4) use of atomic concepts to simplify the understanding of macro- scopic theory of fields inside matter, (5) the relation of’ the micrascopie and macroscopic pictures of electric and magnetic fields inside matter, (6) an introduction to plasma physics, and (7) a substantial list of non- trivial problems which have been carefully related to the textual material. As an aid to the instructor, the more difficult problems have been labeled with an asterisk. Sections of the text which are starred are not essential to its further development, and may be omitted if the course must be shortened for some reason or other. JOR. R. June 1959 PF. J. M.Starred sections and chapters may be omitted without loss of continuity CONTENTS Carrer 1. Vector ANALYSIS... . es == 1-3 1-4. 1-5 1-6 1-7 Definitions .... 2... Vector algebra... . . fe Gradient 22. Se Veetor integration Divergence Cul, boo ooe "Further developments... coo on Carter 2, Exectrosratics 21 22 2-3 24 25 26 27 2-8 29 Electric charge... GG aocucccca: The electric field... . The electrostatic potential . Conductors and insulators... . Gaus law... Application of Gauss’ law The electric dipole... . Multipole expansion of electric fields Cuarter 3. Sorution or Execrrostatic PropLems 3-1 3-2 3-3 34 38 36 *3-7 *3.8 3-9 Poisson’s equation an Laplace’s equation... 2... a Laplace's equation in one independent variable... . Solutions to Laplace’s equation in spherical coordinates. Zonal harmonics : Conducting sphere in a uniform electric field . Cylindrical harmonies... 2... Laplace’s equation in rectangular coordinates Laplace's equation in two dimensions, General solution Electrostatic images 3-10 Point charge and conducting sphere 3-11 Line charges and line images . . . 3-12 System of conductors. Coefficients of potential 3-13 Solutions of Poisson’s equation... . . . vii Foon be 2 21 2 27 30 31 34 36 39 44 44 45 AT 48 bi 52 53 5 56 59 61 63 64viii CONTENTS. Cuarter 4. Tue Evectrosratic Frevp iv Drevecrric Mepia 4-5 z 7 g 4-9 4} Polarization... 2... Soo dc External field of a dielectric medium... . po oos The clectric field inside a dielectric ©... 2... Gouse’ law in a dielectric. ‘The electric displacement . Electric susceptibility and dielectric coustaut 2.2... Point charge in a dielectric fluid . Boundary conditions on the field vectors 2... 2.2... Boundary-value problems involving diclectries . 2 0... . Dielectric sphere in a uniform electric field 4-10 Force on a point charge embedded in a dielectric . . Cuaprer 5. Microscopic Turory or Dreecrrics 1 5-2 *5-3 45-4 Molecular field in a dielectric Induced dipoles. A simple model Polar inolecules. The Langevin-Debye _ Permanent polarization. Ferroelectricity Cuarrer 6. Furcrrosratic ENERGY ge 6-2 6-3 64 $6 6-8 *6-9 Cuaprer 7 ie 7-2 Potential energy of a group of point charges... 2... 1. Electrostatic energy of a charge distribution... .. 2... Energy density of an electrostatic field. . 2... 7 Energy of a system of charged conductors. Coefficients _ tential Coefficients of capacitance and induction Capacitors . Forces and torques . Force on a charge | ‘Thermodynamic interpretation of el costatie Snergy RIG CURRENT Nature of the current . : Current density. Equation of continui iy Ohm's Law. Conductivity Resistance networks Electromotive force Steady currents in media without s sources of emf Approach to electrostatic eq Kirchhoft’s laws Metallic conduction 105, 105 106 109 133 136 139 140 142CONTENTS Cuarren 8. Tur Maeneric Freup or Steavy Currenrs 8-1 The definition of maguetie induction 8-2 Forces on current-carrying conductors . . &-3 The law of Biotand Savart 2... ee ee 8-4 Eleinentary applications of the Biot ond Savart law 8-5 Ampere’s circuitel law 8-6 The magnetic vector pote bocd &-7 The magnetic eld of « distant cirenit . 8-8 ‘The magnete scalar potential 8-9 Magneticfux ... . . ars - Cuarrer 9. Exrctromacsetic Inpuction 9-1 Electromagnetic induction . . . 9-2 Selfinductanee . . 9-8 Mutualinductance . . Od The Neumann formula 9-5 Inductances in series and in parallel Csapren 10. Magnetic Paorertizs or Marrer 10-1 Magnetization... 16-2. The mognetic eld produced by ‘uaguetived material 10-3 Maguetic scalar pots 10-4 Sources of the magret: 10-5 The field equations 16-8 Magnetic susceptibility 16-7 Boundary conditions on the field vectors 10-8 reuits containing magnetic media 10-8 netic circuits. 6. Sogo 10-16 Magnetic circuits containing permanent magnets . . j0-11 Boundary-value problems involving magnetic materials *Cpyaprer 11. Muicroscoric TREOR’ oF THE Magnetic Puovarrins oy Marter 5 Molecuisr field inside matter Origin of diamnagnetism Origin of paramagnetism Theory of ferromagnetism . Ferromagnetic domains Ferrites . “152 165x CONTENTS Carrer. 12. Magnetic Enexey 12-1 12-2 12-3 1244 Cuaprer 13. Stowny VARYING CURRENTS . . =e 13-2 13-3 13-4 13-5 13-6 13-7 13-8 13-9 13-10 13-11 *CuarTer 14. Puystcs or Puasmas 77 14-2 14-3 44 14-5 14-6 14-7 Magnetic energy of coupled circuits . . Energy density in the magnetic field Forces and torques on rigid circuits Hysteresis loss . Introduction... 2. ‘Transient and steady-state behavior Rrchhotelews) (000608 or Elementary transient behavior . Steady-state bebavior of a simple series cireuit Sories and parallel connection of impedances . . Power and power factors... . en Resonance 565500 Mutual inductances in ae circuits... Mesh and nodal equations... . Driving point and transfer impedances Electrical neutrality in a plasma Particle orbits and drift motion in a plasma Magnetic mirrors. 2... ‘The hydromagnetic equations The pinch effect... Plasma oscillations and wave motion ‘The use of probes for plasma measurements Carter 15. Maxwet’s Equations . . . 15-1 15-2 15-3 15-4 15-5 15-6 ‘15-7 15-8 The generalization of Ampere’s law. Displacement current Maxwell’s equations and their empirical basis Electromagnetic energy... 20... The wave equation... . Plane monochromatic waves in nonconducting media Plane monochromatic waves in conducting media . . Spherical waves 2... 2. The wave equation with sources Cuaprer 16. APPLICATIONS OF MAXWELL’s Equations 16-1 Boundary conditions... .. 16-2 Reflection and refraction at the boundary of two nonconducting media. Normal incidence... .. 231 231 235 238 244 244 245 246 247 251 252 254 255 2357 261 265 269 271 272 277 278 281 284 287 301*16-3 16-4 16-5 16-6 16-7 16-8 16-9 CONTENTS Reflection and refraction at the boundary of two nonconducting media. Oblique incidence... ... Reflection from a conducting plane. Normal incidence Propagation between parallel conducting plates 2... .. Waveguides 2. ee Le Cavity resonators oe ee Radiation from an oscillating dipole... . . Radiation from a half-wave antenna Carter 17, ExLecrropyNAMics 17-1 17-2 17-3, 17-4 Appenpix I, Logical Definition of mks Units APPENDIX II, Other Systems of Units A Answers To Opp-NUMBERED PRoBLEMS InpEx The Lienard-Wiechert potentials . . The field of a uniformly moving point charge Radiation from an accelerated point charge Radiation fields for small velocities . . . enprx IIL. Proof that div B = 0 and curlB = poJ 327 332 334 343 344 348 351 351 353 357 360 365 367 369 375 383CHAPTER 1 VECTOR ANALYSIS In the study of electricity and magnetism a great saving in complexity of notation may be-accomplished by using the notation of vector analysis. In providing this valuable shorthand, vector analysis also brings to the forefront the physical ideas involved in equations. It is the purpose of this chapter to give a brief but self-contained exposition of basic vector analysis and to provide the rather utilitarian knowledge of the field which is required for a treatment of electricity and magnetism. ‘Those already familiar with vector analysis will find it a useful review and an introduction to the notation of the text. -1-1 Definitions. In the study of elementary physics several kinds of quantities have been encountered; in particular, the division into vectors and scalars has been made. For-our purposes it will be sufficient to define a scalar as follows: A scalar is a quantity which is completely characterized by its magnitude, Examples of scalars are numerous: mass, time, volume, etc. A simple extension of the idea of a scalar is a scalar field, i.e., a function of position which is completely specified by its magnitude at all points in space. A vector may be defined as follows: A vector is a quantity which is completely characterized by its magnitude and direction. As examples of vectors we cite position from a fixed origin, velocity, acceleration, force, etc. The generalization to a vector field gives a func- tion of position which is completely specified by its magnitude and direo- tion at all points in space. These definitions may be refined and extended; in fact, in more advanced treatments they are usually replaced by more subtle definitions in terms of transformation properties. In addition, more complicated kinds of quantities, such as tensors, are sometimes encountered. Scalars and vectors will, however, suffice for our purposes. 1-2 Vector algebra. Since the algebra of scalars is familiar to the reader, this algcbra will be used to develop vector algebra. In order to proceed with this development it is convenient to have a representation of vectors, for which purpose we introduce a three-dimensional cartesian 12 VECTOR ANALYSIS {cnap. 1 coordinate system. This three-dimensional system will be denoted by the three variables x, y, 2 or, when it is more convenient, 2, 22, 73. With respect to this coordinate system a vector is specified by its z-, y-, and z-components. Thus # vector* V is specified by its components Vz, Vy, Ve, where Vz = [V| cos a1, Vy = |V| cosas, Vz = [V| cosas, the a's being the angles between V and appropriate coordinate axes. In the case of vector fields, each of the components is to be regarded as a function of x, y, and z, It should be emphasized at this point that we introduce a representation of the vectors with respect to a cartesian coordinate system only for simplicity and ease of understanding; all of the definitions and operations are, in fact, independent of any special choice of coordinates. ‘The sum of two vectors is defined as the vector whose components are the sums of the corresponding components of the original vectors. ‘Thus if C is the sum of A and B, we write C=A+B (1-1) and Cr= Art Bz Cy= Ayt By, Cr = Av +B. (1-2) This definition of the vector sum is completely equivalent to the familiar parallelogram rule for vector addition. Vector subtraction is defined in terms of the negative of a vector, which is the vector whose components are the negatives of the corresponding components of the original vector. Thus if Ais a vector, —A is defined by (-A)e= —An (“Aly = Ay (“Ale = — Ae (1-8) ‘The operation of subtraction is then defined as the addition of the negative. This is written A—B=A-4- (—B). (1-4) Since the addition of real numbers is associative, it follows that vector addition (and subtraction) is also associative. In vector notation this appears as A+ B+O = +B)+C=(A+C)+B=A+B+C. ; (1-8) In other words, the parentheses are not needed, as indicated by the last form. Proceeding now to the process of multiplication, we note that the simplest product is a scalar times a vector. This operation results in a * Vector quantities will be denoted by boldface syibols.1-2] VECTOR ALGEBRA 3 vector each component of which is the scalar times the corresponding component of the original vector. If ¢ is a scalar and A a vector, the product. cA is a vector, B = cA, defined by Be= ch, By = chy Br = cA, (1-8) It is clear thei if A is a vector field and ca scalar field then B is a new vector field which is not necessarily a simple multiple of the original field. If, now, two vectors are to be multiplied, there are two possibilities, known as the vector and scalar products. Considering first the scalar product, we note that this name derives from the scalar nature of the prod- uct, although the alternative names, inner product and dot product, are sometimes used. The definition of the scalar product, written A - B, is A-B= A,B,+ A,B, + A:B. (1-7) Thiv definition is equivalent to another, and perhaps more familiar, definition, i.e., as the product of the magnitudes of the original vectors times the cosine of the angle between these vectors. The vector product of two vectors is a vector, which accounts for the name. Alternative names are outer and cross product. The vector product is written A X B; if C is the vector product of A and B, then C = A X B, or Cr = A,B, — A:By, Cy = A,B: — AzB:, = A,By — AyBz. (1-8) ‘This definition is equivalent to the following: the vector product is the product of the magnitudes times the sine of the angle between the original vectors, with the direction given by a right-hand screw rule.* It is important to note that the cross product depends on the order of the factors; interchanging the order introduces a minus sign. ‘The vector product may be easily remembered in terms of a determinant. If i, j, and k are unit vectors, ic., vectors of unit magnitude, in the z-, y+, and z-directions, respectively, then ij k& | AXB=|4, Ay Az (1-9) B, By Bz * Let A be rotated into B through the smallest possible angle. A right-hand screw rotated in this manner will advance in a dircction perpendicular to both A and B; this direction is the direction of A X B.4 VECTOR ANALYSIS (cuar. 1 If this determinant is evaluated by the usual rules, the result is precisely our definition of the cross product. At this point one might well inquire as to the possibility of vector divi- sion. Division of a vector by a scalar can, of course, be defined as multi- plication by the reciprocal of the scalar. Division of a vector by another vector, however, is possible only if the two vectors are parallel. On the other hand, it is possible to write general solutions to vector equations and so accomplish something closely akin to division. Consider the equation c=A-X, (i-10) where c is a known scalar, A a known vector, and X an unknown vector. A general solution to this equation is cA k= 5 t+B (4-11) where B is a vector of arbitrary magnitude which is perpendicular to A, that is, A-B = 0. What we have done is very nearly to divide c by A; more correctly, we have found the general form of the vector X which satisfies Eq. (1-10). There is no unique solution, and this fact accounts for the vector'B. In the same fashion we may consider the vector equation C=AxX, (i-12) where A and C are known vectors and X is an unknown vector. The general solution of this equation is CxA X= 7 + kA, (1-13) where k is an arbitrary scalar. This again is very nearly the quotient of C by A; the scalar & takes account of the nonuniqueness of the process. If X is required to satisfy both (1-10) and (1-12), then the result is unique and given by _ cA atin (1-14) x= re The algebraic cperations discussed above may be combined in many ways. Most of the results so obtained are obvious; however, there are two triple products of sufficient importance to’ merit explicit mei The triple scalar product D = A-B x C is easily found to be given by the determinant i | Az Ay Az| D=A-BxC=|B, B, B,|=—B-AXC. (1-18) Ce Cy Gs |1-3} GRADIENT 5 ‘This product is unchanged by an exchange of dot and cross or by a cyclic permutation of the three vectors; parentheses are not needed, since the eross product of a scelar and a vector is undefined. The other interesting triple product is the triple vector product D = Ax (BX). By a repeated application of the definition of the cross product, Eq. (1-8), we find D=AX(BXC) = BA-C) — C(A-B), (1-16) swhich is frequentiy known as the “back cab rule.” It should be noted that. in the eross product the parentheses are vital; without them the product is not well defined. 1-3 Gradient. The extensions of the ideas intreduced above to differ- entiation and integration, i.e., vector calculus, will now be considered. The simplest of these is the relation of a particular vector field to the derivatives of a scalar field, It is convenient to first introduce the idea of the directional derivative of a function of several variables. This is just the rate of change of the function in a specified direction. The directional derivative of a scalar function ¢ is usually denoted by de/ds; it must be understood that ds represents an infinitesimal displacement in the direc- tion being considered, and that ds is the scalar magnitude of ds. If ds hes the components dz, dy, dz, then La, y ae dy, Oy de ay ds" 32 ds In order to clarify the idea of a directional derivative, consider a scalar function of two variables. Thus, ¢(z,y) represents a two-dimensional sealar field. We may plot gas a function of x and y as is done in Fig, 1-1 for the function g(x,y) = z? 4 y. The directional derivative at the point xo, yo depends on the direction. If we choose the direction corre- sponding to dy/dx == —.ro/yo, then we find dy) _ dg de, ap dy _f, Ey Aslecyo Oe det dy de ~ [270 ~ Wor, Alternatively, if we choose dy/dx = yo/to, we find —— we) |__ 2 ay i. 2x9 + ao Vat = 2VaF +93. (I-17b) dg ds6 VECTOR ANALYSIS [cnar. 1 Fie. 1-1. The function ¢(z, y) = 22+ y? plotted against x and y ina three- dimensional graph. As a third possibility choose dy/dz = a; then fe) = xy + 2ayo)(L + a2)72?2, (1-17) 8 |eo.v0 If this result is differentiated with respect to « and the derivative set equal to zero, then the value of a for which the derivative is a maximum or minimum is found. When we perform these operations, we obtain a = yo/to, which simply means that the direction of maximum rate of change of the function ¢ = z? + y? is the radial direction. If the direc- tion is radially outward then the maximum is the maximum rate of in- crease; if it is radially inward it is a maximum rate of decrease or minimum rate of increase. In the direction specified by dy/de = —zo/yo the rate of change cf 2? + y? is zero. This direction is tangent to the circle x+y? = 23+ y2. Clearly, on this curve, ¢ = 2? + y? does not change. The direction in which dg/ds vanishes gives the direction of the curve g = constant through the point being considered. ‘These lines, which are cireles for the function x? + y?, are completely analogous to the familiar contour lines or lines of constant altitude which appear on topographic maps. Figure 1-2 shows the function g = 2? 4- y? replotted as a contour map.1-3] GRADIENT 7 Fic. 1-2. The function g(z, y) of Fig. 1-1 expressed as a contour map in two dimensions. The idea of contour lines may be generalized to a function of three variables, in which case the surfaces, (zx, y, 2) = constant, are called level surfaces or equipotential surfaces. The three-dimensional analog to Fig. 1-2 is the only practical way of graphing a scalar field for a three- dimensional space. The gradient of a scalar function may now be defined as follows: The gradient of a scalar function g is a vector whose magnitude is the maximum directional derivative at the point being considered and whose direction is the direction of the maximum directional derivative at the point. It is evident that the gradient has the direction normal to the level surface of ¢ through the point being considered. The most common symbols for the gradient are V and grad; of these we will most often use the latter. In terms of the gradient the directional derivative is given by de _ cE = lgrad o| cos 6, (1-18) where @ is the angle between the direction of ds and the direction of the gradient. This is immediately evident from the geometry of Fig, 1-3. If we write ds for the vector displacement of magnitude ds, then (1-18) can be written de 8, & . grade (1-19) This equation enables us to find the explicit form of the gradient in any coordinate system in which we know the form of ds. In rectangular coor- dinates we know that ds = idx + jdy + kdz. We also know that8 VECTOR ANALYSIS [cuap. 1 Ws 2). |grad gl Fio, 1-3. Parts of two level surinces of the fun is the corresponding at P equals the limit as PQ -+ 0 of Ag/P@, and dy/ limit of Ag, = Le 88 4, 1 OF do = rae + 54 au + 3 ae. mm this and Hq. (1-19) it follows that ae oy Equating coefficients of differentials of independent variables on both sides of the equation gives we + (grade), dy (grad 0), de. 88 ety eed 5p + == (grad w grado (1-20) in rectangular coordinates. In a more coi the same. In spherical polar coordinates, with r, 6, pas we have and ds = a, dr + aer dé +- agrsin ¢ dq, (1-22) where a,, a, and a are unit vectors in the r, #, and ¢ directions respectively. Applying (1-19) and equating coefficients of independent variables yields =a eta 44, ¥ “2 grade = a + ae ag + Mersin O6 (1-23) in spherical covrdinates. 1-4 Vector integration, There are, of course, other aspecis to differen- tiation involving vectors; however, it is convenient to discuss vector1-4] VECTOR INTEGRATION 9 Fic. 1-4. Definition of the polar coordinates 7, 8, b= integration first. For our purposes we may consider three kinds of inte- grals: line, surface, and volume, according to the nature of the differential appearing in the integral. The integrand may be either vector or a -scalar; however, certain combinations of integrands and differentials give rise to uninteresting integrals. Those of interest here are the Line integral of a vector, the surface integral of a vector, and the volume integrals of both vectors and scalars. Tf F is a vector, the line integral of F is written. f F-al, (1-24) ae where © is the curve along which the integration is performed, a and b the initial and Gna} points on thé curve, and di an infinitesimal vector displacement slong the curve C. Since ¥ - di is 8 scalar it is clear that the Hine integral is a scalar. The definition of the lire integral follows closely the Riemaun definition of the definite integral. The segment of C between a and b is divided into a large number of small increments Al,; for each increment an interior point is chosen and the value of F at that point found. The sealar product of each increment with the corresponding value of F is found and the sum of these computed. The line integral is then defined as the limit of this sum as the number of increments becomes infinite in such a way that each increment goes to zero. This definition may be compactly written as : N f F-dl= lim })F,- Ab. ag Nevo £4 It is important to note that the line integral usually depends nct only on the endpoints a and b but also on the curve C along which the integration10 VECTOR ANALYSIS fewar. 1 is to be done. The line integral around a closed curve is of sufficient importance that a special notation is used for it, namely, $F a. (1-25) The integral around a closed curve may or may not be zero; the class of veetors for which the line integral around any closed curve is zero is of considerable importance. For this reason one often encounters line inte- grals around undesignated closed paths, e.g., ¢ F- dl. (1-26) This notation is useful only in those cases where the integral is inde- pendent of the contour ( within rather wide limits. If any ambiguity is possible, it is wise to specify the contour. The basic approach to the evaluation of line integrals is to obtain a one-parameter description of the curve and then use this description to express the line integral as the sum of three ordinary one-dimensional integrals. In all but the simplest cases this is a long and tedious procedure; fortunately, however, it is seldom necessary to evaluate the integrals in this fashion. As will be seen later, it is often possible to convert the line integral into a more tractable surface integral or to show that it does not depend on the path between the endpoints. In the latter case a simple path may be chosen to simplify the integration. If F is again a vector, the surface integral of F is written f, F-nda, (1-27) where S is the surface over which the integration is to be performed, da is an infinitesimal area on S and n is a unit normal to da. There is a two- fold ambiguity in the choice of n, which is resolved by taking n to be the outward drawn normal if S is a closed surface. If S is not closed and is finite then it has a boundary, and the sense of the normal is important ony with respect to the arbitrary positive sense of traversing the boundary. The positive sense of the normal is the direction in which a right-hand “screw would advance if rotated in the direction of the positive sense on the bounding curve. This is illustrated in Fig. 1-5. The surface integral of F over a closed surface S is sometimes denotéd by $F nda. Comments exactly parallel to those made for the line integral can be made for the surface integral. The surface integral is clearly a scalar; it1-5} DIVERGENCE i usually depends on the surface S, and cases where it does not are particu- larly important. The definition of the surface integral is made in a way exactly comparable to that of the line integral. The detailed formulation is left as an exercise. If F is a vector and ¢ a scalar then the two volume integrals in which we are interested are J= fred, K= f Pav (1-28) Clearly J is a scalar ‘and K a vector. ‘The definitions of these integrals reduce quickly to just the Riemann integral in three dimensions except that in K one must note that there is one integral for each component of F. These integrals are sufficiently familiar to require no further comment. Boundary Fig. 1-5. Relation of normal n to a surface, and the direction of traversal of the boundary. 1-5 Divergence. Another important operator, which is essentially a derivative, is the divergence operator. The divergence of vector F, written div F, is defined as follows: The divergence of a vector is the limit of ils surface integral per unit volume as the volume enclosed by the surface goes to zero. That is, : aa div F = tin, fF maa: The divergence is clearly a scalar point function (scalar field), and it is defined at the limit point of the surface of integration. The above defini- tion has several virtues: it is independent of any special choice of coor- dinate system, and it may be used to find the explicit form of the divergence operator in any particular coordinate.system. In rectangular coordinates the volume element Az Ay Az provides a convenient basis for finding the explicit form of the divergence. If one corner of the rectangular parallelépiped is at the point xo, yo, zo, then12 VECTOR ANALYSIS fonsr. 1 Fi(2o + Ax, y, 2) = Faia, ys 2) + A. F(x, yo + dv, 2) = Fylx, wo; 2) + Ay (2-29) Hs \ F(t, y 29 + Ae) = F(a, y, 20) + 42 oF 2 Neys0! where higher-order terms in Az, 4y, and Az have been omitted. Since the area element Ay Az is perpendicular’to the x-axis, Az Ar perpendicular to the y-axis, and Az Sy perpendisular to the z-axis,.the definition of the divergence becomes divF = iu { por ¥, #) dy dz OP, , fame s “+ Ag Ay aos + fre Ho 2) de de ; [rec y) 20) dx dy + Ar dy a2 Et f Bela, y, 2) dy de r ; ~~ J Pe, wos 2) di de — | Fae, ws 29) de dy} (1-30) ‘The minus signs associated with the last three terms account for the fact that the outward drawu normal is in the direction of the negative axes in these cases. The limit is easily taken, and the divergence in rectangular coordinates is found to be OF, , Fy 5 OF az ay div F = (1-31) In spherical coordinates the procedure is similar. The volume enclosed by the coordinate intervals Ar, A9, Ag is chosen as the volume of integra- tion. This volume is r? sia @ Ar A@ Ag. Because the area enclosed by the coordinate intervals depends or the values of the coordinates (note that this is not the ease with rectangular coordinates), it is best to write F -n Az in its explicit form: Fenda = Fy? sin 640 Ad +, Por sin @ Ag Ar + Far Ar AG. (1-32) It is clear from this expression that r7f, sin 6, cather than just F,, must be expanded in Taylor series. Similarly, it is the coc‘ficient of the products1-5) GIVERGENCE B of coordinate istervals which must be expanded in the other terms. Mak- ing these expansions and using ther to evaluate the surface integral in the definition of the divergence gives fa 4 lar (By sin 0 Ar 6 A@ div F = lim os 9 on, ag a +h Ss (Myr sin 8) M0 Ar Ag + os (Pst) Ab Ar aa}. (1-33) Talsing the limit, the explicit form of the divergenee in spherical coordi- nates is found to be & fein 6F 6) + > div F - (1-34) rsind d@ ‘This saethod of Enciing the explicit form of the divergence 1 applicable to any coordinate aystem provided the forms of the velume and surface elemente or, alternatively. the elements of ienyth are know. ‘The physical significance of the divergence is readily seen in terms of an example taken from fiuid mechanics. If W is the velocity of a fluid, given as u function of position, and p is its density, then fy pV «nda is clearly the net amount of, laid per unit time that leaves the volume enclosed by 8. If the fluid is incompressible, the surface invesral measures the total source of fluid enclosed by the surface. The above definition of the diver- gence then indioates that it may be interpreted as the limit of the source strength per unit volume, or the source density of an incompressible fluid. An éxtremely important the nvolving the divergence may now be stated and proved. — The integral of the divergence of a vector over a ia the surface integral of the normal component of the vecior over the surface bounding V. That is, {div Fav b Pada, ivided into a large uumber of small ceils. and be bounded by the surface S; It Consider the volume to be sul Lot the ith celt have volume AV, ix clear that Le ¥ nda ¢, F-uda, é egral cu the leit the normel is mitward from the volume being considered. Since outward to ane cell is iuvard to the appropriate adjacent rell, all contributions to the left side of (1-35) cancelWA VECTOR ANALYSIS [cnar. 1 except those which arise from the surface of S, and Eq. (1-35) is essen- tially proved. The divergence theorem is now obtained by letting the number of cells go to infinity in such a way that the volume of each cell goes to zero. . i 7 fF mde = lim Slab fF nad avi. (1-36) Av;40 In the limit, the sum on 7 becomes an integral over V and the ratio of the integral over S; to AV; becomes the divergence of F. Thus, ¢, Fenda = f, div F dv, (1-37) which is the divergence theorem. We shall have frequent occasion to ex- ploit this theorem, both in the development of the theoretical aspects of electricity and magnetism and for the very practical purpose of evaluating integrals. 1-6 Curl. The third interesting vector differential operator is the curl. The curl of a vector, written curl F, is defined as follows: The curl of a vector is the limit of the ratio of the integral of its cross product with the outward drawn normal, over a closed surface, to the volume en- closed by the surface as the volume goes tw zero. That is, i curl F = lim ¢ nx Fda. (1-38) The parallelism between this definition and the definition of the diver- gence is quite apparent; instead of the scalar product of the vector with the outward drawn normal, one has the vector product. Otherwise the defini- tions are the same. This definition is convenient for finding the explicit form of the curl in various coordinate systems; however, for other purposes a different but equivalent definition is more useful. This alternative definition is: The component of curl F in the direction of the unit vector a is the limit of a line integral per unit area, as the enclosed area goes to zero, this area being perpendicular to a. ‘That is, a-curlF = lim 1g F-di, (1-39) S08 where the curve C, which bounds the surface S, is in a plane normal to a. It is easy to see the equivalence of the two definitions by considering a plane curve C and the volume swept out by this eurve when it is dis- placed a distance £ in the direction of the normal to its plane, as shown in1-6) cURL 1 —_—— Fig. 1-6. Volume ewept out by displacing the plane curve ¢ in the direction of its normal, a. Fig. 1-6. If a is normal to this plane, then taking the dot product of a with the first, definition of the curl, (1-38), gives a-curlF = lim 4¢ a-nx Fda. (40) vo VJs Since a is parallel to the normal for all of the bounding surface except the narrow strip bounded by C and C’, only the integral over this surface need be considered. For this surface we note that a X nda is just édl, where dl is an infinitesimal displacement along C. Since, in addition, V = 88, the limit of the volume integral is just = a-curlF = tin Af a, which reduces to the second form of our definition upon cancelling the #s. This equivalence can be shown without the use of the special volume used here; however, so doing sacrifices much of the simplicity of the proof given above. ‘The form of the cur! in various coordinate systems can be calculated in much the same way as was done with the divergence. In rectangular coordinates the volume Az Ay Az is convenient. For the 2-component of the curl only the faces perpendicular to the y- and z-axes contribute. Recalling that j X k = —k X j =i, the nonvanishing contributions rom the faces of the parallelepiped to the z-component of the curl give (eurl Be = tim (Pte, ype + As) + Fyfe y, 1 A ay a + [F.(x, y + Ay, 2) — Fiz, y, 2) Ar Az}. (1-41) Making a Taylor series expansion and taking the limit gives (curl F),= e - of (1-42)16 VECTOR ANALYSIS louap. 1 for the x-component of the curl. The y- and z-components may be found in exactly the'same way. They are af, oF dz ox {curl F), == 1 url), = 8 S. a) ‘The form of the curl in rectangular coordinates can be easily remembered if it is noted that it is just the expansion of a three-by-three determinant, namely, i | a @ % Er (4) Fy FP, The problem of finding the form of the curl in other coordinate systems is only slightly more complicated and is left to the exercises. ‘As with the divergence, we encounter an important and useful theorem involving the curl, known as Stokes’ theorem. Sroxss’ Tunorem. The line integral of a vector around a closed curve is equal lo the integral of the normal component of its curl over any surface bounded by the curve, That is, $, F-d= f, curl F- nda, (1-45) where’ C is a closed curve which bounds the surface S. The proof of this theorem is quite analogous to the proof of the divergence theorem. The surface S is divided into a large number of cells. The surface of the ith cell is called AS; and the curve bounding it is C;. Since each of these cells must be traversed in the same sense, it is clear that the sum of the line integrals around the C;’s is just the line integral around the bounding curve; all of the other contributions cancel. Thus $F d= Ue dl. (146) It remains only to take the limit as the number of cells becomes infinite in such a way that the area of each goes to zero. The result of this limiting process is : 1 fra sin, Daf Faas: I, curl F- nda, (+47)1-7] FURTHER DEVELOPMENTS 7 which is Stokes’ theorem. This theorem, like the divergence theorem, is useful both in the development of electromagnetic theory and in the evaluation of integrals. It is perhaps worth noting that both the diver- gence theorem and Stokes’ theorem are essentially partial integrations. 1-7 Further developments. The operations of taking the gradient, divergence, or curl of appropriate kinds of fields may be repeated. For example, it makes sense to take the divergence of the gradient of a scalar field. Some of these repeated operations give zero for any well-behaved field. One is of sufficient importance to have a special name; the others can be expressed in terms of simpler operations. An important double operation is the divergence of the gradient of a scalar field. This combined operator is known as the Laplacian operator and is usually written V? In rectangular coordinates, 2 ao, Py , de Ve = se tage tar (1-48) This operator is of great importance in electrostatics and will be con- sidered at length in Chapter 3. The curl of the gradient of any scalar field is zero. This statement is most easily verified by writing it out in rectangular coordinates. If the scalar field is ¢, then ijk _{a @ a}]_ fae ie) 7 curl grad ¢ = —_. =1(#¢ - fe +++:=0, (1-49) ae de a9 1x dy dz which verifies the original statement. The divergence of any curl is also zero. This is verified directly in rectangular coordinates by writing a (aF, _ aF, _ 4 )+2(& —St)4---=0. (1-50) oz oz ox The other possible second-order operation is taking the curl of the curl of a vector field. It is left as an exercise to show that in rectangular eo- ordinates curl curl F = grad div F — v°F, (1-51) where the Laplacian of & vector is the vector whose rectangular compo- nents ate the Laplaciens of the rectangular components of the original vector. In any coordinate system other than rectangular the Laplacian -of a veotor is defined by Eq. (1-51).18 VECTOR ANALYSIS fouar. 1 Another way in which the application of the vector differential opera- tors may be extended is to apply them to verious products of vectors and scalars. There are many possible combinations of differential operators and products; those of most interest are tabulated in Table 1-1. These identities may be readily verified in rectangular coordinates, which is sufficient to assure their validity in any coordinate system. Tassie 1-1 Formunas From Vector ANALysIs [NVOLVING Dirrerentiau OPERa tors . : | GD) vet) = vet w | (1-2) Yoh = ow + Ve | (L-3) div (F + G) = divF + divG (1-4) curl (F + G) curl F -+ curl G | (I-5) V@-G) = (F-V)G+ (G-V)F+F XK ariG+ GX curl F (I-6) div oF = pdivF+F-Vo | (-7) div (FX G) = G-eurl F — F-curiG | (L8) diveulF = 0 (L-9) curlgF = geurlF + We xX F (1-10) curl (F X G) = FdivG — GdivF + (G-¥)F — &-¥)G | (I-11) carl curl F grad div F — v?F (1-42) curl ve = 0 (I-13) $F nda = fy divF dv (I-14) foF-dh = fgcurlF-ada (15) Ssynda = fy Ve du * (I-16) fs F(G-n) da = fy F div G dv + Jy (G+ OF dv (-17) fn X Fda = fy curl F dv (I-18) Seed = fen X Veda There are several possibilities for the extension of the divergence theorem and of Stokes’ theorem. The most interesting of these is Green’s theorem, which is I, W820 — oWy) dv ¢, (¥ grad » — ggrady)-nda. (1-52) ‘This theorem follows from the application of the divergence theorem to the vector P grad —- y grad y. (1-53) Using this F in the divergence theorem, we obtainPROBLEMS 19 j, div [y grad y — ygrad yj] dv = $, (grad y -- o grad) -nda. (1-54) Using the identity (Table 1-1) for the divergence of a scalar times a veetor gives div (Y grad y) — div (y grad y) = WW?e — eV*¥. (1-55) Combining (1-54) and (1-55) yields Green’s theorem. This conciudes our brief discussion of vector analysis, In the interests of brevity, many well-known results have been relegated to the exercises. Ne attempt has been made to achieve a high degree of rigor; the approach has been utilitarian, What we wili need we have developed; everything alse has heen omitted. PROBLEMS 1-1, The vectors from the origin to the points 4, B,C, D are =it+it+k = 2+ 3i, i+ 5j — %, ak ~j. Show that the lines AB and CD are parollel and find the ratio of their lengths. 1-2. Show that the fotlowing vectors are’ perpendieniar: A sit 4j-+ 3k, B= 4i-b 23 — 4k, 1-3, Show that the vectors fi power A=a—i+k, B =i 3j — 5k, Cc = 3i — 4j — 4k form the sides ofa right: triangle. }-4, By squaring both sides of the equation A=B—C and interpreting the result geometrically, prove the “iaw of cosines.” 1-5. Show that A = icosa + jsina, B = icos@-+ jsinB are unit vectors in the zy-plane making angles «, 8 with the a-axis, By means of a scalar product, obtain the formule for cos (e — 8).20 VECTOR ANALYSIS [cuap. 1 1-6. If A is a constant vector and r is the veetor from the origin to the point (x, y, 2), show that (@-A)-A=0 is the equation of a plane. 1-7. With A and r defined as in Problem 1-6, show that (@— Ayr =0 is the equation of a ephere. 1.8. If A, B, C, are vectors from the origin to the points A, B, C, show that (AX B)+ (BX C)+ (CX A) is perpendicular to the plane ABC. 1-9. Verify that Eq. (1-13) is a solution to (1-12) by direct substitution. {Note that Eq. (1-12) implies that C is perpendicular to A.] 1-10. Find the gradient of ¢ in cylindrical coordinates, given that ds = dra, + r day + dek. It should be noted that r and @ have different meanings here than in Eqs. (1-21) and (J--22). In spherical coordinates r is the magnitude of the radius vector from the origin and @ is the polar angle. In cylindrical coordi- nates, r is tho perpendicular distance from the cylinder axis and 9 is the azimuthal angle about this axis. 1-11. From the definition of the divergence, obtain an expression for div F in cylindrical coordinates. 1-12. Find the divergence of the vector A(x? + ye) + jy? + 22) + (2? + ay). Also find the curl. 1-13. If ris the vector from the origin to the point (x, y, 2), prove the formulas divr = curlr = 0; (u-grad)r =u. (Note: u is any veotor.] I-L4. 1f A is a constant veetor, show that grad (A-r) = A. 1-15. Prove identities (I-6) and (1-8) in Table 1-1. 1-16. If r is the magnitude of the vector from the origin to the point (z, y, 2), and f(r) is an arbitrary function of r, prove that a r dr 1-17. Verify Eq. (1-61) in rectangular coordinates, where V2F in these coordinates is as defined in the text. 1-18. Prove identities (I-15) and (I-16) in Table 1-1. (Hint: Use the diver- gence theorem and one or more identities from the first half of Table 1-1.] grad f(r) =CHAPTER 2 ELECTROSTATICS 2-1 Electric charge. The first observation of the electrification of ob- jects by rubbing is lost in antiquity; however, it is common experience that rubbing 9 hard rubber comb on a piece of wool endows the rubber with the ability to pick up small pieces of paper. As a result of rubbing the two objects together (strictly speaking, as a result of bringing them into close contact), both the rubber and the wool acquire a new property; they are charged. This experiment serves to introduce.the concept of charge. But charge, itself, is not created during this process; the total charge, or the sum of the charges on the two bodies, is still the same as-before elec- trification. In the light of modern physics we know that microscopic charged particles, specifically electrons, are transferred from the wool to the rubber, leaving the wool positively charged and the rubber comb negatively charged. Charge is a fundamental and characteristic property of the elementary particles which make up matter. In fact, all matter is composed ultimately of protons, neutrons, and electrons, and two of these particles bear charges. But even though on a microscopic scale matter is composed of a large number of charged particles, the powerful electrical forces associated with these particles are fairly well hidden in a macroscopic observation. The reason is that there are two kinds of charge, positive and negative, and an ordinary piece of matter contains approximately equal amounts of each kind. From the macroscopic viewpoint, then, charge refers to nct charge, or excess chargé. When we say that an object is charged, we mean that it has an excess charge, either an excess of electrons (negative) or an excess of protons (positive). In this and the following chapters, charge will usually be denoted by the symbol g. Since charge is a fundamental property of the ultimate particles making up matter, the total charge of a closed system cannot change. From the macroscopic point of view charges may be regrouped and combined in different ways; nevertheless, we may state that net charge is conserved in a closed system. 2-2 Coulomb’s law. Towards the end of the eighteenth century tech- niques in experimental science achieved sufficient sophistication to make fdssible refined observations of the forces between electric charges. ‘The results of these observations, which were extremely controversial at the time, can be summarized in three statements. (a) There are two and only 2122 ELECTROSTATICS {cnar. 2 two kinds of electric charge, now known as positive and negative. (b) Two point charges exert on each other forces which act along the line joining them and which are inversely proportions! to the square of the distance between them. {c) These forces are also proportional to the product of the charges, are repulsive for like charges, and attractive for unlike charges. The last two statements, with the first as preamble, are known as Coulomb's law in honor of Chatles Augustin de Coulomb (1736-1806), who was one of the leading eighteenth century students of electricity. Coulomb’s law for point charges may be concisely formulated in the vector notation of Chapter i as — =O 2e tn F=C ee (2-1) where F; is the force on charge qi, Tai is the vector from gz to q1, 721 is the magnitude of r2;, and C is a constant of proportionality about which more will be said later. In Eq. (2-1) a unit vector in the direction of tz has been formed by dividing rz, by its magnitude, a device of which frequent use will be made. If the force on gz is to be found, it is only necessary to change every subscript 1 to 2 and every 2 to4. Understanding this notation is important, since in future work it will provide a tech- nique for keeping track of field and source variables. Coulomb’s law applies to point charges. In the macroscopic sense a “point charge” is one whose spatial dimensions are very small compared with any other length pertinent to the problema under consideration, and we shall use the term “point charge” in this sense. To the best of our knowledge, Coulomb's law also applies to the interactions of elementary particles such as protons and electrons. Equation (2-1) is found to hold for the electrostatic repulsion between nuclei at distances greater than about 107'* meter; at smaller distances, the powerful, but short-ranged, nuclear forces dominate the picture. Equation (2-1) is an experimental law; nevertheless, there is both theoretical and experimental evidence to indicate that the inverse square law is exact, ie., that the exponent of ra; is exactly 2. By an indirect experiment’ it has been shown that the exponent of re: can differ from 2 by no more than one part in 10%. 7 ‘The constant C in Eq. (2-1) requires some comment, since it determines the system of units. The units of force and distance are presumably those belonging to one of the systems used in mechanics; the most direct procedure here would be to set C = 1, and choose the unit of charge such that Eq. (2-1) agrees with experiment. Other procedures are also " * Plimpton and Lawton, Phys. Rev. $0, 1066 (1936). The same experiment: was performed earlier by Kelvin and by Maxwell. Maxwell established the exponent of 2 to within one part in 20,000.~2-2} COULOMR’S LAW 23, possible and ruay have certain advaninges; ¢.g., the unit of charge may be specified in advance, 4 was shown by Giorgi in 1901 that all of the common electrival units, such as the ampere, volt, ohm, hei can be combined with one of the mechanical systems (name or meter-kilogram-second system) to form o system of units for ali elec- trie and magnetic problems. There is considerable advantage tu having the results of calculations come out in the serae units as those which are used in the laboratory; hence we hall use the r A mks or Giorgi system of units in the present volume. Since in this system ¢ is meusured in coulorabs, r in meters and F in newtons, it is ciear that C must have the dimensions of newton-meters?/coulomb?. ‘The size of the unit of charge, the coulomb, is established from “magnsti ments; this requires that’ C = 8.9874 x 10° n-m7/coul?. We make the apparently complicated substitution, C = 1/47¢o, in the interest of feture simplicity. The constant €9 will occur repeatedly; it represents s property of free space known as the permittivity of free space, and is numerically equal to: 8.854 X 107*? coul?/n-m?. In Appendix I the otinitions of the coulomb, the ampere, the permeability, and permittivity f free space aro slated to one another and to the velocity of light in a logical way; since a logical formulation of these definitions requires a Imowledge cf magnetic phenomena and of electromagnetic wave prc agation, it is not appropriate to pursue them now. In Appendix IT othe systems of electrical units, in particular the gaussian system, ure disensced. If more then two point charges are present, the mutual forces are determined by the repeated ‘application of Eq. (2-1). In particular, if » system of N charges is considered, the foree on the ith ehatge is given by ) where the summation on the right is extended over all of the churges except the ith. This is, of course, just the superposition principle for forces, which says that the total force acting on a body is the vector sum of the individual forces which act on it. A simple extension of the ideas of N interacting point charges is the interaction of a point charge with a continuous charge distribution. We deliberately choose this configuration to avoid certain difficulties which may be encountered when the interaction of two continuous charge dis- tributions is considered. Before proceeding further the meaning of a con- tinuous distribution of charge should be examined. It is now well known that electric charge is ‘found in multiples of a basic charge, that of the electron. In other words, if any charge were examined in great detail, its magnitude would be found to be an integral multiple of the magnitude of the electronic charge. For the purposes of macroscopic physics this24 ELECTROSTATICS: [cnar. 2 discreteness of charge causes no difficulties simply because the electronic charge has a magnitude of 1.6019 x 107? coul, which is extremely small. The smallness of the basic unit means that macroscopic charges are in- variably composed of a very large number of electronic charges; this in turn means that in a macroscopic charge distribution any small element of volume contains a large number of electrons. One may then describe a charge distribution in terms. of a charge density function defined as the limit of the charge per unit volume as the volume becomes infinitesimal. Care must be used, however, in applying this kind of description to atomic problems, since in these cases only a small number of electrons is involved, and the process of taking the limit is meaningless. Leaving aside these atomic cases,‘we may proceed as if a segment of charge might be sub- divided indefinitely, and describe the charge distribution by means of point functions: a volume charge density defined by = 2-3) em bm op C3 and a surface charge density defined by: Aq a= Jim 55° (2-4) From what has been said about q, it is evident that p aud o are net charge, or excess charge, densities. It is worth while mentioning that in typical solid materials even a very large charge density p will involve a change in the local electron density of only about one part in 10°. If charge is distributed through a volume V’ with a density p, and on the surface S which bounds V with a density ¢, then the force exerted by this charge distribution on-a point charge q located at r is obtained from (2-2) by replacing g; with p, dv; (or with 9 da’) and proceeding to the limit: r—r “ Gees vir— SB ar) do! + +e al, jer ACE a2) The variable r’ is used to locate a point within the charge distribution, that is, it plays the role of the source point r; in Eq. (2-2). It may appear at first sight that if point r falls inside the charge distribution, the first integral of (2-5) should diverge. This is not the case; the region of inte~ gration in the vicinity of r contributes a negligible amount, and the integral is well behaved (see Problem 2-5). It is clear that the force on g as given by Eq. (2-5) is proportional to g; the same is true in Eq, (2-2). This observation leads ys, tp introduce a vector field which is independent of g, namely, the force per unit charge.2-3] ‘THE ELECTRIC FIELD 25 ‘This vector field, known as the electric field, is considered in detail in the following section. 2-3 The electric field. The electric field at a point is defined as the limit of the following ratio: the force on a test charge placed at the point, to the charge of the test charge, the limit being taken as the magnitude of the test charge goes to zero. The customary symbol for the electric field is E. In vector notation the definition of E becomes E = lim £e. (2-6) a0 7 ‘The limiting process is included in the definition of E to ensure that the test charge does not affect the charge distribution which produces E. If, for example, positive charge is distributed on the surface of a conductor (a conductor is a material in which charge is free to move), then bringing a test charge into the vicinity of the conductor will cause the charge on the conductor to redistribute itself. If the electric field were calculated using the ratio of force to charge for a finite test charge, the field obtained vovid be that due to the redistributed charge rather than that due to the al charge ribution. In the special case where one of the charges of the charge distribution can be used as a test charge the limiting process ts unnecessary. In this case the electric field at the location of the test charge will be that produced by all of the rest of the charge distribution; therevill, of course, be no redistribution of charge, since the proper charge distribution obtains under the influence of the entire charge distribu- tion, including the charge being used as test charge. In certain other cases, notably those in which the charge distribution is specified, the force will be proportional to the size of the test charge. In these cases, too, the limit is unnecessary; however, if any doubt exists, it is always safe to use the limiting process. Equations (2-2) and (2-5) provide a ready means for obtaining an ex- pression for the electric field due to a given distribution of charge. Let the charge distribution consist of N point charges q1, g2,-..., gv located at the points r1, Ta, . Ty respectively, and a volume distribution of charge specified by the charge density p(r’) in the volume V and a surface distri- bution characterized by the surface charge density o(r’) on the surface S. If a test charge q is located.at the point r, it experiences a force F given by i. + Gre [ q r tre Js — 8" a o(r’) da’, (2-7)26 ELECTROSTATICS lonar. 2 due to the given charge distribution, The electric field at r is the limit of the ratic of this force to the test charge q. Since the ratio is independent field at r is just j ott") do’ _ a(t’) da’, (2-8) rad reg Js It ib) Fig. 2-1 The mapping of aa elcetric field with tie aid of lines of force:24} THE ELECTROSTATIC POTENTIAL 7 Equation (2-8) is very general; in most cases one or more of the terms will not be needed. The quantity we have just defined, the électric field, may be calculated at each point in space in the vicinity of a system of charges or of a charge distribution. Thus E = E(r) is a vector point function, or a vector field. ‘This field has a number of interesting mathematical properties which we shall proceed to develop in the following sections and in the next chapter. As an aid to visualizing the electric field structure associated with a particular distribution. of charge, Michael Faraday (1791-1867) intro- duced the concept of lines of force. A line of force is an imaginary line (or curve) drawn in such a way that its direction at any point is the direction of the electric field at that point. Consider, for example, the electric. field structure associated with a single positive point charge q. ‘The lines of force are radial lines radiating outward from qi. Similarly, the lines of force associated with an isolated negative point charge are also radial lines, but this time the direction is inward , toward the negative charge). These two examples are ex- tremely simple, but they nevertheless illustrate an important property of the field lines: the lines of force terminate on the sdurces of the electric field, i-e., upon the charges which produce the electric field. Figure 2-1 shows several simple electric fields which have been mapped with the aid of lines of force. 2-4 The electrostatic potential. It, has been noted in Chapter 1 that if the curl of a vector vanishes, then the vector may be expressed as the gradient of a scalar. The electric field given by Eq. (2-8) satisfies this cviterion. To verify this, we note that taking the curl of Eq. (2-8) involves differentiating with respect to r. ‘This variable appears in the equation only in functions of the form (x — r')/|r — 2'|3, and hence it will suffice to show that functions of this form have zero curl. Using the formula from Table 1-1 for the curl of the product (vector times scalar) gives ) 4 [grad od ie = 2 curl (r ~ r’) + jerad Fo ri] Xf}. (3-9) A direct calculation (see Problem 1-13) shows that curl (r — r’) = 0, (2-10) and (see Problem 1-16) that (2-11)28 ELECTROSTATICS [cnar. 2 These results, together with the observation that the vector product of a vector with a parallel vector is zero, suffice to prove that Y 7 = 2 curl a = 0. (2-12) Since each contribution of Eq. (2-8) to the electric field is of this form, we have demonstrated that the curl of the electric field is zero. Equa- tion (2-12) indicates that a scalar function exists whose gradient is the electric field; it remains to find this function. That is, we now know that a function exists which satisfies E(r) = —grad U(r), (2-13) but we have yet to find the form of the function U. It should be noted that it is conventional to include the minus sign in Eq. (2-13) and to call U the electrostatic potential. Tt is easy to find the electrostatic potential due to a point charge qi; it is just u@ =e Treo ’ (2214) as is readily verified by direct differentiation. With this as a clue it is easy to guess that the potential which gives the electric field of Eq. (2-8) is 1) = a4 tf ©) ay UO} = are 2a — alt Gree Jy o : 2) da’, (2-15) ire Js FFI which is also easily verified by direct differentiation, It may seem that Eqs. (2-14) and (2-15) were obtained in 2 rather arbitrary fashion; how- ever, since all that is required of U is that ii satisfy (2-13), and since this has been verified directly, the means by which U was obtained is im- material. The electrostatic potential U can be obtained directly as soon as its existence is established. Since U is known to exist, we may write iE E(r’) dr’ = — f. grad U- dr’, (2-16) tof ref where ref stands for a reference-point at which U is zero. From the defi- nition of the gradient, grad U-dr' = dv. (2-17) Using (2-17) in Eq. (2-16) converts it into the integral of a perfect dif-2-4] THE ELECTROSTATIC POTENTIAL 29 ferential, which is easily done. The result is - ik grad U- dr’ = —U(r) = f. E(r’)- dr’, (2-18) rot ret which is really the inverse of Eq. (2-13). If the electric field due to a point charge is used in equation (2-18), and the reference point or lower limit in the integral is taken at infinity, with the potential there zero, the result is . vw) =a (2-19) This, of course, is just a special case of Eq. (2-14), namely, the case where 11 is zero. This derivation can be extended to obtain Eq. (2-15); however, the procedure is too cumbersome to include here. Another interesting and useful aspect of the electrostatic potential is its close relation to the potential energy associated with the conservative electrostatic force. The potential energy associated with an arbitrary conservative force is WG) = — [_ Fe) -ae, (2-20) where W(r) is the potential energy at r relative to the reference point at which the potential energy is arbitrarily taken to be zero. Since in the electrostatic case F = gE, it follows that if the same reference point is chosen for the electrostatic potential and for the potential energy, then the electrostatic potential is just the potential energy per unit charge. This idea is sometimes used to introduce the electrostatic potential; we feel, however, that the introduction by means cf Eq. (2-13) emphasizes the importance of the electrostatic potential in determining the electro- static field. There is, of course, no question about the ultimate equivalence of the two approaches. The utility of the electrostatic potential in calculating electric fields can be seen by contrasting Eqs. (2-8) and (2-15). Equation (2-8) is a vector equation; to obtain the electric field from it, it is necessary to evaluate three sums or three integrals for each term. At best this is a tedious procedure; in some cases it is almost impossible to do the integrals. Equation (2-15), on the other hand; is a scalar equation and involves only one sum or integral per term. Furthermore, the denominators appear- ing in this equation are all of the ferm |r — rj, which simplifies the inte- grals compared with those of Eq. (2-8). This simplification is sometimes sufficient to make the difference between doing the integrals and not doing them. It may be objected that after doing the integrals of Eq. (2-15) it is still necessary to differentiate the result; this objection is readily answered by observing that differentiation can always be accomplished30 ELECTROSTATICS [cuar. 2 if the derivatives exist, and is in fact usually much easier than integration. Yn Chapter 3 it will be seen that the electrostatic potential is even more important in those problems where the charge distribution is not specified, but must rather be determined in the process of solving the problem. In the mks system the unit of energy is the newton-meter or joule. The unit of potential is joule/coulomb, but this unit occurs so frequently that it is given a special name, the volt. The unit of the electric field is the newton/coulomb or the volt/meter. 2-5 Conductors and insulators. So far as their electrical behavior is concerned, materials may be divided into two categories: conductors of electricity and insulators (dielectrics). Conductors are substances, like the metals, which contain large numbers of essentially free charge carriers. ‘These charge carriers (electrons in most cases) are free to wander through- out the conducting material; they respond to almost infinitesimal electric fields, and they continue to move as long as they experience a field. These free carriers carry the electric current when an electric field is maintained in the conductor by an external source of energy. Dielectrics are substances in which ali charged particles are bound rather strongly to constituent molecules. The charged particles may shift their positions slightly in response to an electric field, but they do not leave the vicinity of their molecules. Strictiy speaking, this definition applies to an ideal dielectric, one which shows no conductivity in the presence of an externally maintained electric field. Real physical dielec- tries muy show.a feeble conductivity, but in a typical dielectric the con- ductivity is 10° times smaller than that of a good conductor. Since 107° is a tremendous factor, it is usually sufficient to say that dielectrics are nonconductors. Certain materials (semiconductors, electrolytes) have electrical prop- erties intermediate between conductors and dielectrics. So far as their behavior in a static electric field is concerned, these materials behave very much like conductors. However, their transient response is somewhat slower; i.e., it takes longer for these materials to reach equilibrium in 4 static field. : In this and the following four chapters we shall be concerned with materials in electrostatic elds. Dielectric polarization, although a basically simple phenomenon, produces some rather complicated effects; hence we shall delay its study until Chapter 4. Conductors, on the other hand, may be treated quite easily in terms of concepts which have already been developed. Since charge is free to move in a conductor, even under the influence of very small electric fields, the charge carriers (electrons or ions) move until they find positions in which they experience no net force. When2-6] GAUSS’ LAW 31 they come to rest, the interior of the conductor must be a region devoid of an electric feld; this must be so because the charge carrier population in the interior is by 20 means depleted, and if a field persisted, the carriers would continue to move. Thus, under static conditions, the electric field in a conductor vanishes. Furthermore, since £ = 0 in a conductor, the tential is the same at all points in the conducting material. Tn other words, under static conditions, each conductor forms an eyuipotential region of space. 2-6 Gauss’ iaw. An important relationship exists between the integral of the normal component of the electric field over a closed surface and the total charge enclosed by the surface. This relaticaship, kaown as Gauss’- jax, will now be investigated in more detail. ‘Phe electric field at point due to & point charge gq located at the origin is EQ) = (2-21) Consider the surface integral of the normal component of this electric field over a closed surface (such as that shown in Fig. 2-2) which encloses and, consequently, the charge q; this integral is just _G gta 2! fe. adam eh a (2-22), Fig, 2-2, An imaginary closed surface $ which encloses a point charge at the origin. The quantity (r/r) - nda is the. projection of da on a plane perpendicular tor. This projected area divided by r? is the solid angle subtended by da, which is written dQ It is clear from Fig. 2-3 that the solid angle sub- tended by do-is the same as the solid angle subtended by da’, an element ef the surface area of thé sphere S’ whose center is at the origin and whose radius is r’. It is then possible to write Se a=.32 ELECTROSTATICS {cnar. 2 Fie. 2-8. Construction of the spherical surface S’ as an aid to evaluation of the solid angle subtended by da. which sbows‘that | 2 2 - po nde = Gear = a (2-23), in the special ease described above. If q lies outside of S, it is clear from Fig. 2-4 that S can be divided into two areas S,-and Sz each of which subtends the same solid angle at the charge g. For Sa, however, the direction of the normal is towards g, while for S, it is away from q. ‘There- fore the contributions of S, and S» to the surface integral are equal and opposite, and the total integral vanishes. Thus if the surface surrounds a poine charge g, the surface integral of the normal component of the electric field is o/éo, while if q lies outside the surface the surface integral is zero. n Fic. 2-4, The closed surface S may be divided into two surfaces, $1 and Sa, cacti of which subtend the same solid angle at ¢. The‘preceding statement applies to any closed surface, even to so-called re-entrant ones. A study of Fig. 2-5 is sufficient to verify that this is. indeed the case. 7 If several point charges 91, gz, ..., gv are enclosed by the surface S, then the total electric field is given by the first term of Eq. (2-8). Each2-6] Gauss’ LAW 33 Fia. 2-5. An element of solid angle cutting the surface S more than once. charge subtends a full solid angle (47); hence Eq. (2-23) becomes 1 N fee nda = aut (2-24) This result can be readily generalized to the case of a continuous dis- tribution of charge characterized by a charge density. If each element of charge p dv is considered as a point charge, it contributes p dv/€y to the surface integral of the normal component of the electric field provided it is inside the surface over which we integrate. The total surface integral ig then the sum of all contributions of this form due to the charge inside the surface. Thus if S is a closed surface which bounds the volume V, A tnda=i Y 25; fe nda 2 faa. (2-25) Equations (2-24) and (2-25) are known as Gauss’ law. The term on the left, the integral of the normal component of the electric field over the surface S, is sometimes called the flux of the electric field through S. Gauss’ law may be expressed in yet another form by using the divergence theorem. The divergence theorem (1-37) states that ¢ F-nda =f div F do. s v If this theorem is applied to the surface integral of the normal component of E, it yields ¢ E-nda= f div E dv, (2-26) s v which, when substituted into Eq. (2-25), gives . 1 f div Edy = 2 fae (2-27)4 ELECTROSTATICS: [ouiar. 2 Equation (2-27) must be valid for all volumes, that is, for any choice of the volume V. ‘The only way in which this can be true is if the integrands appearing on the left and on the rigbt in the equation are equal. Thus the validity of Eq. (2-27) for any choive of V implies that (2-28) This result may be thought of as 4 differential form of Gauss’ law. 2-7 Application of Gauss’ law. Equation (2-28) or, more properly, a modified form of this equation which will be derived in Chapter 4, is one of the basic differe itial equations of electricity and magnetism. {n this role it is important, of course; but Gauss’ law aiso has. practical utility. This practicality of the law lies largely in providing a very easy way to calculate electric fields in sufficiently syrametric situations. In other words, in certain highly symmetric situations of considerable physical interest, the electric field may be calculated by using Gauss’ law instead of by the integrals given above or by the procedures of Chapter 3. When this can be done, it accomplishes a major saving in effort. Part of long | line charge Fic. 2-6. A cylindrical surfaee to be used with Cause’ law to find the electric field produced by a long line charge. In order that Gauss’ law be useful in calculating the electric field, it must be possible to choose « closed surface such that the electric field has a normal component which is either zero or @ single fixed value at every. point on the surface. As an example, consider 2 very long line charge of charge density ) per unit length, as shown in Fig. 2-6. The symmetry of2-7) APPLICATION OF GAUSS’ LAW 35, the situation clearly indicates that the electric field is radial and inde- pendent of both position along the wire and angular position ardund the wire. These observations lead us to choose the surface shown in Fig. 2-6. For this surface it is easy to evaluate the integral of the normsi component of the electrie field. The circular ends contribute nothing, since the electric field is parallel to them. ‘The cylindrical surface contributes 2rriZi, since E is radial and independent of the position of the cylindrical surface. Gauss’ law then takes the form aang, = (2-29) £0 Equation (2-29) can be solved for B, to give a trey (2-30) The saving of effort accomplished by the use of Gauss’ law will be more fully appreciated by solving Problem 2-4, which involves direct applica tion of Eq. (2-8). . Another important result of Gauss’ law is that the charge (net charge) of a charged conductor resides on its surface. We saw in Section 2-5 that the electric field inside a conductor vanishes. We may construct a gaussian surface anywhere inside the conductor; by Gauss’ law, the net charge enclosed by each of these surfaces is zero. Finally, we construct the gaussian surface S of Fig. 2-7; again the net charge enclosed is zero. The only place left for the charge. which is not in contradiction with Gauss’ law is for it to reside on the Surface of the conductor. Conductor Fic. 2-7, A gaussian surface $ constructed inside a charged conductor. The electric field just outside a charged conductor must be normal! to the surface of the conductor. This follows because the surface is an equi- potential, and E = ~grad U. Let us assume that the charge on a con- ductor is.given by the surface density function ¢. If Gauss’ law is applied to the:small pillhox-shaped surface S of Fig. 2-8, then BAS = (2) AS,36 ELECTROSTATICS fenar. 2 Fic. 2-8. Application of Gauss’ law to the closed, pillbox-shaped surface 'S which intersects the surface of a charged conductor. avhere AS is the area of one of the pillbox bases. Hence, for the electric field just outside a conductor, o E= oo (2-31) 2-8 The electric dipole. Two equal and opposite charges separated by a small distance form an electric dipole. The electric field and potential distribution produced by such a charge configuration can be investigated with the aid of the formulas of Sections 2-3 and 2-4. Suppose that a charge —q is located at the point r’ and a charge q is located at r’ + J, as shown in Fig. 2-9; then the electric field at an arbitrary point r may be found by direct application of Eq. (2-8). The electric field at r is found to be vy g-f{r-r—t orn) 2®) = Galkor i ~ Fore (2-82) ‘This is the correct electric field.for any value of q and any value of the separation J; however, it is not easy to interpret. What we want is the dipole field, and in the dipole the separation J is small compared with 1 — 1’; hence we may expand Eq. (2-32), keeping only the first non- vanishing term. Since this procedure is of general utility it will be con- sidered in detail. The primary difficulty in making this expansion is caused by the denominator of the first term of Eq. (2-32). The reciprocal: of this denominator can be rewritten as fe — rf — 4 = (@ — v)? — 2 — 9) 1 PS? Qr —r’)-l meer — EE2-8] THE ELECTRIC DIPOLE 37 ro 0 Fic. 2-9. Geometry involved in calculating the electric field E(r) due to two point charges. In the last form it is easy to expand by the binomial theorem, keeping only terms linear in 7, The result of this expansion is Br) +l rrp jp—-r — YP = fr — vrfi +o fee |. (2-33) where terms involving 1? have been dropped. Using Eq. (2-33) in Eq. (2-32) and again keeping only terms linear in J gives ioe pean: t fae Ue 78 Equation (2-34) gives that part of the electric field, due to a finite electric dipole, which is proportional to the separation of the charges. There are other contributions proportional to the square, the cube, and higher powers of the separation. If, however, the separation is small, these higher powers contribute very little. In the limit .as J goes to zero, all of the terms vanish unless the charge becomes infinite. In the limit as 1 goes to zero while g becomes infinite, in such a way that gl remains constant, all terms except the term linear in 2 vanish. In this limit a point dipole is formed. A point dipole has no net charge, no extent in space, and is completely characterized by its dipole moment, which is the limit of gl as Z goes to zero. We use the symbol p to represent the electric dipole moment, and write p= dl. (2-35) EQ) = (@—r)— : | (2-34) In terms of, the dipole moment, Eq. (2-34) may be written EQ) = 1 SE=P) Pe — py trea PE (2-36)38 ELECTROSTATICS [ewar. 2 The potential distribution produced by a point dipole is alse important. This could be found by looking for a function with gradient equal to the right side of Eq. (2-36). It is, however, easier to apply Eq. (2-14) to the charge distribution consisting of two point charges separated by a small distance. Using the notation of Eq. (2-32), the potential distribution is given by U@)= | rr (2-37) By expanding the first. term in exactly the same way as was done for the first term of (2-32) and retaining only the linear term in t, Eq. (2-37) can be put in, the form U@ =7e Grey (2-38) This equation is valid to the same approximation as Eq. (2-34); namely, terms proportional to l# and to higher powers of I are neglected. For a point dipole, p, Eq. (2-38) is exact; however, it is better written as (2-89) Equation (2-39) gives the potential U(r) produced by an electric dipole; from this potential the electric field (Eq. 2-36) may be determined. It is also interesting to inquire about the potential energy of an electric dipole which is placed in an external electric field. In the case of two charges, —qatrand qatr -+ J, in an electric field described by the potential fune- tion Uext(r), the potential energy is just, W = —qQenl(t) + qestlt +). (2-40) If Lis small compared with 1, User(t -+ 1) may be expanded in a power series in 7 and only the first two terms kept. The expansion gives Uex(e + D = Uexe(t) + 1+ grad Vext, (2-41) where the value of the gradient at point r is to be used. Ii this expansion is used in Eq. (2-40), the result is W = q+ grad Use (2-42) Going to the limit of a point dipole gives simply W(r) = pr grad Ue, (2-43) which is, of course, exact. Since the electric field is the negative gradient2-9) MULTIPOLE EXPANSION GF ELECTRIC FIELDS 39 of the electrostatic potential, an alternative form of Eq. (2-43) is WO) = —p- Eoult) (2-44) This, then, is the potential energy of @ dipole p in an external electric field Een, where Eex(t) is evaluated at the location of the dipole. It is important to note that two potentials have been discussed in’ this section. In Eqs. (2-37), (2-38), and (2-29), the electrostatic potential produced by an electric dipole is considered. In Eqs. (2-40) through (2-43), the dipole is considered to be in an existing electric field described by a potentiai function Usx(r). ‘This electric field is due to charges other than those comprising the dipole; in fact, the dipole field must be excluded to avoid an infinite result. ‘This statement could lead us to rather complicated questions concerning sc!f-forces and self-energies, which we cannot discuss here; however, it may be noted that the potential energy resulting from the interaction of an electric dipole with its own field arises from forces exerted on the dipole by itself. Such forces, known in dynamies as internal forces, do not affect the motion of the dipole as a whole. Fer our purposes further consideration of this question will be unnecessary. 2-9 Multipole expansion of electric fields. It is apparent from the defmition of dipole moments given above that certain aspects of the potential distribution produced by a specified distribution of charge might well be expressed in terms of its electric dipole moment. In order to do this it is necessary, of course, te define the electric dipole moment of an arbitrary charge distribution. Rather than make an unmotivated defini- tion, we shall consider a certain expansion of the elecirostatic potential due to an arbitrary charge distribution. To reduce the number of position coordinates, a charge distribution in the neighborhood of the origin of ecordinates will be considered. The further restriction will be made that the charge distribution can be entirely enclosed by a sphere of radius « which is sifall compared with the distance to the point of observation. An arbitrary point within the charge distribution will be designated by r’, the charge density at that point by e(r’), and the observation point by r (see Fig. 2-10). The potential at r is given by (2-45) where di’ is used to designate an element of volurac iu the charge distribu- tion and 7 denoics the entire voiume occupied uy the charge distribu- . In view of the restriction rade ebave to points of observation which are remote from the origia, the quantity {¢ — ry"! ean be expanded in a series of ascending powers of x’/r. The result of such an expansion is40 ELECTROSTATICS [cuar. 2 Observation point Fig. 2-10. The charge is localized in the volume V with charge density p(t’). The electric field is to be calculated at point r. 1 (7? — Ore et $2)? 1 ajo oon a 113 0. | 1 Yea cag gee re} eam . 2 r re i where only the first three terms are explicitly indicated. It should be noted that while (r’/r)? is negligible compared with 2r’-r/r?, it may not be dropped in the first set of brackets because it is of the same order as the dominant term in the second set of brackets. Using Eq. (2-46) in Eq. (2-45) and omitting terms involving the cube and higher powers of r’ yields t py? wt UG) = asl +E t 1 [ae 0)* = “| 7 to do’. (2-47) Since r does not involve the variable of integration 1’, all of the r depend- ence may be taken from under the integral sign, to obtain um = { f , p(r") de! + f rp(r’) do! Gre, ao 1 xij + Po fede | _ 2 a y ! 2 yy saa [ , (Baits —~ 8477’?)p(a") do’, (2-48) where 2;, 2; are cartesian components of r, xj, x} are the cartesian com- ponents of 1’, and 4,; is defined as follows: It is easy to interpret Eq. (2-48). The first integral in the equation is clearly the total charge, and the first term is the potential which would2-9} MULTIPOLE EXPANSION OF ELECPRIC FIELOS 41 result if this total charge were concentrated ai the origin. The second integral is very similar to the dipole moment. so in Section 2-7, and 69 it is called the dipole momeut of the c! definition, this represents a coe ae of. the “definitio two equal and opposite point charges; ver, that both definitions give the same result for “two © erjual al 9 ‘ite pulnt charges. The second term in Eq. (2-48) is the otentiat which would result if a point dipole equal to the dipole moment « distribu- tion were located at the origin of coordinates. It that the dipole moment of a charge distribution i erdent of the origin of coordinaigs if the total charge is zero. ‘fo verify this, consider a new coordinate system with origin at R in the old system. Denvting a point with respect to the old system by r’ and the same point with respect to the new system by r”, we have =r +R. (2-49) ‘The dipole moment with respect to the old system ie 2 = fi rele) de = fa" + Rp) a! = f pd’: RQ, (2-50) whieh proveg;the statement above, ‘The third term of Eq. (2-48) can be written VY, m1? where Q,; is given by Os = f, Br There are nine components of Q,; corresponding to 7, j equ Of these nine components six are equal in pairs, leaving six distinct com- ponents. This set of quantities forra the quadrupole moment tensor and ‘epresent an extension of the dipole moment concept. There are, of course, igher-order moments which are generated by keeping higher-order terms in the expansion of Eq. (2-48). These higher-order multipcles are im- portant in nuclear physics, but will not be considered further in this back. ‘The electric multipoles are used, a8 Eq, (2-48) indicates, to spproximate the electric field of a charge distribution. There are, however, many other uses, ali in the framework of approximating 4 real extended charge distribution by point charges, puint dipoles, ete. ‘These appro: tions often make it possible to solve problems which would otherwise be pro- hibitively difficult. — bir"? )p lr’) de’. (2-82)42 ELECTROSTATICS [cnar. 2 ProsLems 2-1. Two particles, each of mass m and having charge g, are suspended by strings of length i from a common point. Find the angle @ which each string makes with the vertical. 2-2. Two small identical conducting spheres have charges of 2.0 X 10-9 coul and —0.5 X 10-° coul, respectively. When they are placed 4 em apart, what is the force between them? [f they are brought into contact and then separated by 4 cm, what is the force between them? 2-3. Point charges of 3 X 10~® coul are situated at each of three corners of a square whose side is 15cm. Find the magnitude and direction of the electric field at the vacant corner point of the square, 2-4, Given an infinitely long line charge with uniform charge density \ per unit length. Using direct integration, find the electric field at a distance r from the line. 2-5. (a) A cireular disk of radius ® has a uniform surface charge density o. Find the electric ficid at a point on the axis of the disk at a distance z from the plane of the disk. (b) A right circular cylinder of radius R and height Z is oriented along the z-axis. It has a nonuniform volume density of charge given by plz) = + 82 with reference to an origin at the center of the ¢ Find the force on a point charge q placed at the center of the cylinder. 2-6. A thin, conducting, spherical shell af radius & is charged uniformly with total charge Q. By direct integration, find the potential at an arbitrary point (a) inside the shell, (b) outside the shell. 2-7. Two point charges, —q and +-4q, are situated at the origin and at the point (a, 0,0) respectively. At what point along the z-axis does the electric field vanish? {n the z, y-plane, make a plot of the equipotential surface which gocs through the point just referred to. Is this point a true minimum in the potential? 2-8. Show that the U = 0 equipotential surface of the preceding problem is spherica] in shape. What are the coordinates of the center of this sphere? 2-9. Given a right circular cylinder of radius R and length L containing a uniform charge density p. Calculate the electrostatic potential at a point on the cylinder axis but external to the distribution, 2-10. Giver ‘n of space in which the clectric ficld is everywhere directed parallel to the x Prove that the electric ficld is independent of the y- and z-coordinates in this region. If there is no charge in this region, prove that the field is also independent of 2. 2-11. Given that the dielectric strength of air (i.c., the electric field which produces corona) is 3X 10° v/m, what is the highest possible potential of an isolated spherical conductor of radius 10 em? 2-12. A conducting object has a hollow cavity in its interior. If a point charge q is introduced into the cavity, prove that the charge —q is induced on the surface of the cavity. (Use Gauss’ law.) 2-13. The electric field in the atmosphere at the carth’s surface is approxi- mately 200 v/m, directed downward. At 1400 m above the earth’s surface, the electric field in the atmosphere is only 20 v/m, again directed downward. WhatPROBLEMS 43 js the average charge density in the atmosphere below 1400 m? Does this consist predominantly of posilive or negative ious? / 9-14. Two infinite parallel condueting plates are separated . y the distance d! If the plates have uniform charge densities ¢ and —o, respectively, on their inside surfaces, obtain an expression for the clectrie field between the plates. Prove that the electric ficld in the regions external to the plates is zero. [Two charged parallel conducting plates of finite area produce essentially the sare electric field in the region between then, as was found above provided the di- mensions of the plates are large compared with the separation d; such an arrange- ment is called a capacilor (see Chapter 6).] 2-15. A spherical charge distribution has a volume charge density which is a function only of r, the distarice from the center of the distribution. In other words, p = p(r). If p(r) is as given below, determine the electric ficld as a funetion of r. Integrate the result to obtain an expression for the electro- static potential U(r), subject to the restriction that U(s) = (a) p = A/r with A a constant for 0 > r > R; p = Oforr > R (b) p = po (i.e., constant) for 0 > r > R; p= Oforr> R. 2-16, Using Eq. (2-39) for the potential produced by a dipole p, make a plot of the traces of equipotential surfaces in a plane containing the dipole. For convenience, the dipole may be located at the origin. Use the results obtained to sketch in some of the lines of foree. Compare the result with Fig. 2-1. 2-17. (a) Show that the force acting on a dipole p placed in an external electric field Exxt is p+ VEoxt. (b) Show that the torque acting on the dipole in this field is v= 1X [p+ VEext] + p X Eext, where r is the vector distance to the dipole from the point about which the torque is to be measured. ‘The quantity p X Eexe, which is independent of the point about which the torque is computed, is called the turning couple acting on the dipole. 2-18. Three charges are arranged in a lincar array. ‘The charge ~-2¢ is placed at the origin, and two charges, each of +9, are placed at (0, 0, 2) and (0, 0, ~!) respectively. Find a relatively simple expression for the potential U(r) which is valid for distances |r| >> 1. Make a plot of the equipotential surfaces in the x, zzplane.CHAPTER 3 SOLUTION OF ELECTROSTATIC PROBLEMS The solution to an electrostatic problem is straightforward for the case in which the charge distribution is everywhere specified, for then, es we have seen, the potential and electric field are given directly as integrals over this charge distribution: v@ = a (3) @) = ge (6-2) However, many of the problems encountered in practice are not of this type. If the charge distribution is not specified in advance, it may be necessary to determine the electric field first, before the charge distribution can be calculated. For example, an electrostatic problem may involve several conductors, with either the potential or total charge of each con- ductor given, but the distribution of surface charge will not be known in general, and cannot be obtained until a complete solution to the problem is effected. Our aim in this chapter is to develop an alternative approach to electro- static problems, and to accomplish this we first derive the fundamental differential equation which must be satistied by the potential U. For the present we shall disregard problems invelving dielectric bodies; problems of this type will be solved in Chapter 4. 3-1 Poisson’s equation. All of the basic relationships which we shall need here were developed in the preceding chapter. First, we have the differential form of Gauss’ law, dvE =p. (3-8) &5 Furthermore, in a purely electrostatic field, E may be expressed as minus the gradient of the potential U: E = — grad U, Combining (3-3) and (3-4), we obtain div grad U = -- £. (35a) 0 443-2] LAPLACE’S EQUATION 45 Ji is convenient to think of div grad as a single differential operator, v-¥ or V?. The latter notation is preferred, and the operator is called the Laplacian: vu= —2£. (3-5b) €0 It is evident that the Laplacian is a pure scalar differential operator, and 5b) is a differential equation. This is Poisson's equation. The operator involves differentiation with respect to more than one variable; hence Poisson’s equation is a partial differential equation which may be solved once we know the functional dependene of p(x, y, z) and the appropriate boundary conditions. The operator V®, just like the grad, div, and curl, makes no reference to any particular inate svstem. Tn order to solve @ specific problem, we must write V? in terms of x, y, 2 or 7, #, @, or etc. The choice of the particular set of coordinates is atbitrary, but substantial simplification of the problem is usually achieved by choosing a set compatible with the symmetry of the electrostatic problem. The form taken by V?U in various coordinate systems is easily found by first taking the gradient of U, and then operating with div, using specific expressions from Chapter 1: Rectangular coordinates: (3-6) Spherical coordinates: ee a) 1 3 (s aru. US a ae ar) | ein 8 36 G7) Cylindrical coordinates 2 oi 3 (28 wat Sm) B-8) For the form of the Laplacian in other, more complicated coordinate systems, the reader is referred to the references at the end of this chapter. Jt should be noted that r and have different meanings in (3-7) and (3-8); in spherical coordinates r is the magnitude of the radius vector from the origin and @ is the polar angle. tn cylindrical coordinates, r is the per- pendicalar distance from the cylinder axis and 0 is the azimuthal angle about this axis. 3-2 Laplace’s equation. In a certain class of electrostatic problems involving conductors, all of the charge is found either on the surface of the condvietors or in the form of fixed point charges. In these cases p is46 SOLUTION OF ELECTROSTATIC PROBLEMS [enar. 3 zero at, most points in space. And where the charge density vanishes, the Poisson equation reduces to the simpler form vu = 0, (3-9) which is Larplace’s equation. Suppose we have a set of N conductors (one or more of which may be point charges) maintained at the potentials Uz, Urt,..., Un. Our prob- Jem is to find the potential at all points in space outside of the conductors. This may be accomplished by finding a solution to Laplace’s equation which reduces to Us, Uri,..., Ux on the surfaces of the appropriate conductors. Such a solution to Laplace’s equation may be shown to be unique, i.e., there is no other solution to Laplace’s equation which satisfies the same boundary conditions. A proof of this statement will be given below. The solution to Laplace’s equation which we find in this way is not applicable to the interior of the conductors, because the conductors have surface charge, and this implies a discontinuity in the gradient of U across the surface (see Section 2-7). But we have already seen that the interior of each conductor is a region of constant potential, so the solution to our problem is complete. We shall describe in some detail two methods for solution of Laplace’s equation: the first is a method for compounding a general solution to (3-9) from particular solutions in a coordinate system dictated by the sym- metry of the problem; the second is the method of images. In addition, a completely general solution to the problem in two dimensions will be found. Before taking up these specific procedures, however, we stop to prove some important properties of the solution to Laplace’s equation. Tavorem I. If Us, Uz,..., Un are all solutions of Laplace’s equation, then, U = OU, + 02Ue +e + Cay (3-10) where the €’s are arbitrary constants, is also a solution. The proof of this follows immediately from the fact that ve = WCU, + VC. = O,9°U,; = 0. ee bh VOW MaW*Ug bo + OnV?Un ‘Through the use of Theorem J we may superimpose two or more solutions of Laplace’s equation in such a way that the resulting solution satisfies a given set of boundary conditions, Examples will be given in the following sections,3-3] LAPLACE’S EQUATION IN ONE INDEPENDENT VARIABLE 47 ‘Tusorem II (Uniqueness theorem). Two solutions of Laplace’s equa- tion which satisfy the same boundary conditions differ at most by an additive constant. To prove this theorem we consider the closed region Vo exterior to the surfaces Si, Si1,..., Sw of the various conductors in the problem and bounded on the outside by a surface S, the latter being either a surface at infinity or a real physical surface which encloses Vo. Let us assume that U, and Uz are two solutions of Laplace’s equation in Vo which, in addition, have the same boundary’ conditions on S, St, Six, ..., Sw. These boundary conditions may be specified by assigning values of either U or aU/an on the bounding surfaces, We define a new function 6 = U,; — Us. Obviously, V6 = V2U, — VU, = 0 in Vo. Furthermore, either @ or n- grad & vanishes on the boundaries. Let us apply the divergence theorem to the vector V@: if div (6V®) dV evd-ndS vo Poases- ay =0, since the second integral vanishes. The divergence may be expanded according to Eq. (I-6) of Table 1-1 to give diy (6V®) = 8V°@ + (Vo)?, But V?@ vanishes at all points in Vo, so that the divergence theorem reduces in this case to if (va)? dV = 0. Vo Now (V#)? must be either positive or zero at each point in Vo, and since its integral is zero, it is evident that (V®)? = 0 is the only possibility. The theorem is essentially proved. A function whose gradient is zero at all points cannot change; hence at all points in Vo, @ has the same value that it has on the bounding surfaces. If the boundary conditions have been given by specifying U, and Uz on the surfaces S, S;,..., Sy, then since 6 = 0 on these surfaces, it vanishes throughout Vo. If the boundary conditions are given in terms of 0U,/dn and dU2/dn, then Ve equals zero at all points in Vo and V@-n = 0 on the boundaries. The only solution compatible with the last statement is @ equal to a constant. 3-3 Laplace’s equation in one independent variable. If U is a function of one variable only, Laplace’s equation reduces to an ordinary differential equation. Consider the case where U is U(x), a function of the single48 SOLUTION OF ELECTROSTATIC PROBLEMS [omar. 3 reetangular coordinate x. Then &u dx? is the general solution, where a and b are constants chosen to fit the boundary conditions. This is the result already found in the preceding chapter for the potential between two charged conducting plates oriented normal to the z-axis. ‘The situation is no more complicated in other coordinate systems where U is a function of a single variable. In spherical coordinates where U equals U(r), Laplace’s equation and its general solution become =0 and Ue) =ar+b (+11) 1 d(2 av) __@ pala) =% UN=—F+b. (3-12) The general solution to Laplace’s equation in cylindrical coordinates for a function which is independent of 0 and z, thet is, for U(r), is left as an exercise for the reader. 3-4 Solutions to Laplace’s equation in spherical coordinates. Zonal harmonics. We next turn our attention to solutions of Laplace’s equation where U is a function of more than one variable. Many of the problems of interest, to us deal with conductors in the shape of spheres or cylinders, and thus solutions of Laplace’s equation in either spherical or cylindrical coordinates are called for. We first take up the spherical problem, but we shall find it expedient to limit the discussion to cases in which U is independent of the azimuthal angle ¢. ‘This limitation restricts the class of problems which we shall be able to solve; nevertheless, many interesting physical problems fall into this restricted category, and more complicated probiems are really beyond the scope of this book. For the spherical case, U is U(r, 6), where r is the radius vector from a fixed origin O and @ is the polar angle (see Fig. 3-1). Using Eq. (3-7), Laplace’s equation becomes in this case aa 230) 1a (aU) _ aa Pt 2¢r ar) + ain 6 a6 \8 8 Gg) =O. @-18) This partial differential equation will be solved by a technique known as “separation of variables.” A solution of the form U(r, @) = Z(r)P(@) is substituted into (3-13), yielding Lr@ alr #) ee (sin | (3-14) r®sin 6 dé do) ~ Note that the partial derivatives have been replaced by total derivatives,3-4) SOLUTIONS TO LAPLACE’S EQUATION 49 Polar direction Fie. 3-1. Location of the point P in terms of the spherical coordinates r, 8, e variable only. Dividing thréugh y r®, we transform (3-14) into since Z aud P’ are each functions of by U(r, 8) and muitiplying through & (8-15) function of @ for all values of 7 and @ is for both functions to be constant. let each side of (3-15) equal k, where k is the “separation constant, Not all values of k necessarily yield solutions which are acceptable on physical grounds. Consider the # equation first: . ad ae oe a gin a (sine + kP 0. (3-18) ‘This is Legendre’s equation, and the only physically acceptable solutions which are defined over the full range of 6, from 0 te 7, correspond to k == n(n + 1), where n is a positive integer. The solution for a particular n will be denoted by P,(8). Solutions of (3-16) for other values of k are ill-behaved in the vicinity of @ = 0 or 6 = m radians, becoming infinite or even undefined at these values of @.* These solutions cannot be made to fit physical boundary conditions and hence ntust be discarded.¢ * The discussion here has been all too brief. The interested reader is referred to more mathematical texts for a detailed treatment of Legeudre’s equation. See, e.g., the book by Margenau and Murphy (p. 61) listed at the end of this chapter. Legendre’s equation is usually written in a different form by substituting = = cos @, and utions are then denoted by P,(x) or Pn(cos 6). } This statement requires some qualification. In some electrostatic problems the regions around 0 and 6 = w may be naturally excluded, for example, by condueting conical surfaces; under these conditions solutions of (3-16) with other values of & could he used. Problems of this type will not be considered here.50" SOLUTION OF ELECTROSTATIC PROBLEMS icuap. 3 The acceptable solutions, the P,,(é), are polynomials in cos 8, and are usually referred to as Legendre polynomials: The first four Legendre functions are given in Table 3-1. It is evident from (3-16) that the P, may he multiplied by any arbitrary constant, We now return to the radial equation 4 (? @) = nln + D2, (-17) where we have used the explicit form of k which gave acceptable @ solu- tions. Inspection of (3-17) shows that two independent solutions are Zy=r and Zy == rth, Solutions of Laplace's equation are obtained as the product Un(r, 6) = Za(r) X Pp(6), where particular care must be exereised to have Z and P correspond to the same value of n. This is mandatory, since both sides of Eq, (3-15) are equal to the same constant, namely, n(m + 1). TABLE 3-1 Lecenpre PoryNomIsALs FoR n = 0, 3,2, AnD 3. 4 (3 cos? @ — 1) 4 (5 cos? 6 — cos 8) As a result of the above discussion we have solved Laplace’s equation in spherical coordinates and have obtained the solutions which are known as zonal harmonics: Un = r"P,(0) or Uy = FPP, (8), (3-18) where P,,(9) is one of the polynomials listed in Table 3-1, and n is a positive integer or zero. The zonal harmonics form a complete set of functions, i.e., a general solution of Laplace’s equation may be constructed as a superposition of these solutions according to Theorem I provided the physical problem shows the appropriate azimuthal symmetry. Several of the zonal harmonics are already well known to us: one of the n = 03-5] CONDUCTING SPHERE IN A UNIFORM ELECTRIC FIELD 51 solutions, namely U = constant, is a trivial solution of Laplace’s equation, valid in any coordinate system; the zonal harmonic r~! is the potential of a point charge; and r~? cos @ is the potential of a dipole. 3-5 Conducting sphere in a uniform electric field. We shall illustrate the usefulness of zonal harmonics for electrostatic problems having spheri- cal symmetry by solving the problem of an uncharged conducting sphere placed in an initially uniform electric field Eo. The lines of a uniform electric field are parallel, but the presence of the conductor alters the field in such a way that the ficld lines strike the surface of the conductor, which is an equipotential surface, normally. If we take the direction of the initially uniform electric field as the polar direction (z-direction), and if we make the origin of our coordinate system coincide with the center of the sphere, then from the symmetry of the problem it is clear that the potential will be independent of azimuthal angle , and may be expressed as a sum of zonal harmonies. The spherical conductor, of radius @, is an equipotential surface; let us denote its potential by Uo. Our problem is to find a solution to La- place’s equation in the region outside the sphere which reduces to Uo ou the sphere itself, and which has the correct limiting form at large stances away. The solution may be formally written as UC, 8) = A, + Cyr? + Aor cos 6 + Cor? cos @ + % Agr?(3 cos? @ ~ 1) + $Cgr~8(3 cos? 9 — 1) +--+, (3-19) where the A’s and C’s are arbitrary constants. At large r, the electric field will be only slightly distorted from its initial form, and the potential will be that appropriate to a uniform electric field. (E(r, 6)]p4 = Eo = Eok, (U(r, @)];+2 = —Eye + constant, = —Eor cos 0 + constant. (3-20) Hence, in order to make (3-19) and (3-20) agree at large r, Ay == —Ey; furthermore, all the 4’s from Ag up must be set equal to zero. The term Cyr~ produces a radial field which, as we might expect, is compatible only with a spherical conductor bearing net total charge. Since our problem deals with an uncharged conductor, the Gonstant C, must be set equal to zero, At the surface of the sphere U = Us, and the po- tential must become independent of angle 9. ‘The tw® terms involving cos @ may be made to cancel each other, but the terms with higher inverse Powers of r cannot be cancelled one against the other because they contain52 SOLUTION OF ELECTROSTATIC PROBLEMS lcuAr. 3 ~~ — Fic, 3-2. Lines of electric flux for the case of a conducting sphere placed in a uniform electric field. diferent Legendre functions. The only possibility is to set all the C's with 7 > 3 equal to zero. Equation (3-19) now becomes U(r, 6) = Ay — Eorcos@ + Car-*cos@, for r >a, U(a, @) = Uo. (8-21) Since the two expressions must be equal at r= a, A; = Uo and Cy = Hoa’. From the final expression for the potential, we may calculate not only the electric field at all points in space (see Fig. 3-2) but also the surface density of charge on the conducting sphere: p) = Ba(1 +25) cos ou ot) si ne -}@a—2 (1-3 sin 6, for r>a, (3-22) o(@) = eoE,| = 3€9Ey cos 8. (3-23) ‘The total charge on the sphere, Q=0? f o(6)2m sin 6 d8, ° is obviously zero, which agrees with our initial assumption. 3-6 Cylindrical harmonics. Laplace’s equation in: cylindrical co- ordinates may also be solved by the method of separation of variables. Here again it will be expedient to work out solutions for only a restricted3-7] LAPLACE’S EQUATION IN RECTANGULAR COORDINATES 53 class of problems, namely, those in which the potential is independent of the coordinate z. These solutions are appropriate for certain problenis involving a long straight cylindrical conductor or wire, but not for those dealing with a short cylindrical segment. If the potential is independent of z, Laplace’s equation in cylindrical coordinates becomes 1a av 1 0?U f 7 or (- wv) + a = (B24) Substitution of U = Y(r)S(@) reduces the equation to rd dY 1478 os pa(-2) = -398 (3-25) where k again plays the role of @ separation constant. The @-equation is particularly simple; it has the solutions cos k!/?9 and sin kY/?9. But if these solutions are to make sense physically, each must be a single-valued function of 0; thus cos kY/2(9 + 2m) = cos k¥/6, sin kY?(9 + 27) = sin k!?9, Or, to put it differently, after @ has gone through its full range from 0 to 2m the function must join smoothly to its value at @ = 0. This can-be the case only if k = n?, n being an integer. We may further require n to be positive (or zero) without losing any of these solutions. Returning now to the r-equation, we are able to verify easily that Y(r) is r* or r~*; unless n = 0 when ¥(r) = Inr or Y(r) = constant. Hence the required solutions to Laplace’s equation, the so-called cylindrical harmonics, are 1, Inr, r™cosn8, rr" cos n8, rsinné, "sin nd. These functions form a complete set for the variables r, 6 in cylindrical coordinates, and the potential U(r, @) may be developed as a superposition of cylindrical harmonies in accordance with Theorem 1. *3~7 Laplace’s equation in rectangular coordinates. In rectangular co- ordinates, the variables may be separated by making the substitution U(@, ¥, 2) = fi@felyfale), whereby Laplace's equation reduces to * Starred sections may be omitted without loss of continuity54 SOLUTION OF ELECTROSTATIC PROBLEMS [onar. 3 1 af 1 dfe 1 @fs Fey ae + -- pe (3-26a) ely) dy? ~ The left side of this equation is a function of x and y, and the right side is a function of z only; hence both sides must be equal to the same constant, i. This is the first separation constant. The two equations obtained fron: (3-26a) are @he sy a TE + bs = 0, (3-26) | ‘The latter equation has been written such that the variables # and y are separated; each side of this equation is now set, equal to —m (the second separation constant). Thus, fs Gai + mls = 0, (3-26e) 2 Th & + mf, = 0. (3-264) EQuations (3-26b), (3-26e), and (3-26d) are easily solved. One of the typical solutions for U(x, y, 2) is Ula, yp 2) = Aen *+™*? eos m12y cos bz, (3-27) ‘The ether seven independent solutions for a pair of separation constants (I, m) ure obtained by making one or more of the following substitutions: E(k + m)'!x for — (k + m)"2x, sin my for cos m'?y, and sin k!!?z for cos k!!z, Thus far there are no restrictions on k or m, but boundary conditions on the problem usually restrict & (or m) to a discrete set of positive or negative values. It is worth while making the point that it is the boundary conditions which really pick out the pertinent solutions to a partial differ- ential equation; the function Ube, wy 2) = DD Ange "+ cos py 08 a2 a for fixed x and y is just the Fourier series expansion for an arbitrary even function of z. ‘The individual solutions, (3-27), do not. represent particularly simple potentials, and we shall not try to correlute them with physical situations. The case where both separation constants are zero is more interesting; hence we turn our attention to this case. From (3-264), it is evident that3-8) LAPLACE’S EQUATION IN ‘TWO DIMENSIONS 55 fie) = @12, or f,(x) = constant, is a solution; from (3-26c), we obtain ‘foly), ete. Thus, U(a, y, 2) = Axzyz + Agcy + Agye + Aagzz + Ast t+ Agy + Azz+ Ag, (3-28a) where the A’s are arbitrary constants. This solution may be applied to the case where three conducting planes intersect at right angles. If these planes are the coordinate planes zy, yz, and zz, and are all at the same potential, then UG, y,2) = Arsyz + As. (3-28b) It is left as an exercise for the reader to determine the surface charge density on the coordinate planes that is compatible with (3~-28b). *3-8 Laplace’s equation in two dimensions. General solution. If the potential is a function of only two rectangular coordinates, Laplace’s equation is written (3-29a) It is possible to obtain the general solution to this equation by means of a transformation to a new set of independent variables; nevertheless, it should be emphasized that such a transformation leads to a simplification of the original equation only in the two-dimensional case. Let 2 2a. jy where j = /—1 is the unit imaginary number. In terms of these rela- tionships, a a a = og +? ay? OF and (3-29b) It is evident that the general solution to (3-29b) is U = Fy(f) + Fol) = File + jy) + Fax — jy), (8-30) where F; and F2 are arbitrary functions. The functions 7, and Fz are complex quantities in general, but two real functions may be constructed in the following way. First let Fo(x — jy) = Fy(x — jy), that is, let56 SOLUTION OF ELECTROSTATIC PROBLEMS [cuar. 3 the two functions F, and F2 have the same dependence on their argu- ments; then Uy = File + dy) + Pile — jy) = 2 Re [Fie + jy], where Re stands for “real part of.” Furthermore, the second real potential function is U2 = —jF ila + jy) — File — jy)] = 21m [Fi (2 + jy], where Im stands for “imaginary part of.” Thus the real and imaginary parts of any complex function F(x + jy) are both solutions of Laplace’s equation. ‘The solutions found in this way are not restricted to any particular coordinate system. For example, the cylindrical harmonics of Section 3-7 are obtained from the complex functions* (x + jy)" = r"e’"*, and In (z + jy) = Inr +50. On the other hand, when it comes to solving a particular two-dimensional problem, there is no standard procedure fgr finding the appropriate complex function. This method generates so many solutions that it is not possible to enumerate them all and cast out those which do not agree with boundary conditions on the problem. In simple cases, the required functions may be found by trial and error; in other cases, the method of conformal mapping (which is beyond the scope of this book) may be useful. 3-9 Hlectrostatic images. For a given set of boundary conditions. the solution to Laplace's equation is unique, so that if one obtains a solution U(e,y,2) by any means whatever, and if this U satisfies all boundary conditions, then a complete solution to the problem has been effected. The method of images is a procedure for accomplishing this result without specifically solving a differential equation, It is not universally applicable to all types of electrostatic problems, but enough interesting problems fall into this category to make it worth while discussing the method here. Suppose the potential may be written in the following way’ UG) = Use) + ee I, Oe. (@-31) where U; is either a specified function or easily calculable, and the integral represents the contribution to the potential from surface charge on all conductors appearing in the problem. The function o is not known. It may happen, and this is the essence of the image-charge method, that the last term in (3-31) can be replaced by a potential U2 which is due to * The cylindrical and rectangular coordinates are related in the usual way? xz = rcos8,y = rsin8.3-9) ELECTROSTATIC IMAGES 57 a specified charge distribution. This is possible so long as the surfaces of all conductors coincide with equipotential surfaces of the combined Uy -+ Ug. The specified charges producing U2 are called image charges. They do not really exist, of course. Their apparent location is “inside” the various conductors, and the potential U = U; + U; is a valid solution to the problem only in the exterior region. . As an example of this method, we shall solve the problem of a point charge g placed near a conducting plane of infinite extent. To formulate the problem mathematically, let the conducting plane coincide with the y2-plane, and let the point charge lie on the z-axis at « = d (see Fig. 3-8a). (x, 2) @, 2) Fic. 3-3. Problem of a point charge and conducting plane solved by means of the image-charge method: (a) original problem, (b) location of image charge, (© lines of force (dotted) and equipotential surfaces (solid).38 SOLUTION OF ELECTROSTATIC PROBLEMS [owar. 3 The potential fits the prescription (3-31), with q q 2 Ui@,y,2) =p = a =+ (8-82) 1, y, 2) Gren ~ eV GL OTL Pe (3-32) Consider now a different problem, that of two point charges (g and —g) a distance 2¢ apart, as shown in Fig. 3-3(b). The potential of these two charges, y 4g q Ue?) = aes > age’ (3-33) not only satisfies Laplace’s equation at all points exterior to the charges, but aiso reduces to a constant (namely, zero) on the plane which perpen- dicularly bisects the segment joining the two charges. ‘Thus (3-33) satisfies the boundary conditions of the original problem. Because solutions to Laplace’s equation are unique, (3-33) is the correct potential in the entire half-space exterior to the conducting plane. The charge —. which gives rise to the potential Go q OAD ares — Sree OE is called the image of the point charge g. Naturally, the image does not really. exist, and (3-32) does not give correctly the potential inside or to the left of the conducting plane in Fig. 3-3(a). The electric field E in the exterior region may be obtained as the negative gradient of (3-33). Since the surface of the conducting plane represents an interface joining two solutions of Laplace’s equation, namely, U = and (3-33), the discontinuity in the electric field is accommodated by a surface charge density o on the plane: id oW.2) = Blo = — RET Tm (8S) The lines of force and equipotential surfaces appropriate to the original problem are shown in Fig. 3-3(c). These are the same lines of force and. equipotential surfaces appropriate to the two point charge problem in Fig, 3-3(b) except that in the latter case the flux lines would continue into the left half-plane. It is evident from the figure that all of the electric flux lines which would normally converge on the image charge are inter- cepted by the plane in Fig. 3-3(c). Hence the total charge on the plane is equal to that of the image charge, —g. This same result may be ob- tained mathematically by integrating (3-35) over the entire surface (see Problem 3-10). It is evident that the point charge q exerts an attractive force on the. plane, because the induced surface charge is of the opposite sign. By3-10] POINT CHARGE AND CONDUCTING SPHERE 39 Newton’s law of action and reaction, this force is equal in magnitude to the force exerted on q by the plane. Since the point charge experiences no force due to its own field, F = —qgrad U2, (3-86) which is just the force exerted on it by the image charge. Another problem which may be solved simply in terms of images is that of determining the electric field of a point charge q in the vicinity of a right- angle intersection of two conducting planes (see Fig. 3-4a). ‘The positions of the necessary image charges are shown in Fig. 3-4(b). Tt is readily seen that the two planes shown dotted in the figure are surfaces of zero poten- tial due to the combined potentials of g and the three image charges. qe (a) Fie. . Point charge in a right-angle corner. 3-10 Point charge and conducting sphere. The principal difficulty’ in solving a problem by image technique is that of finding a group of image charges which, together with the originally specified charges, produce equipotential surfaces at the conductors. The problem is straightforward only in cases where the geometry is simple. Such is the case, however, for a point charge q in the vicinity of a.conducting sphere; it requires a single image charge to make the sphere a surface of zero potential. An additional image charge is needed to change the potential of the sphere to some other constant value. We shall first determine the magnitude and location of the i image q which together with the point charge q produces zero potential at all points on the sphere. The geometry of the situation is shown in Fig. 3-5. The point charge q is a distance d from the center of the sphere, and the radius of the sphere is a. It is apparent from the symmetry of the problem that the image charge q’ will lie on the line passing through q and the center of the sphere. The desired results are most easily obtained by means of spherical coordinates, with the origin of coordinates at the center of the sphere.60 SOLUTION OF ELECTROSTATIC PROBLEMS [cnar. 3 P(r, 0, 4) Fic. 3-5. Point charge g in the vicinity of a condneting sphere; ¢' is the image charge. Let the polar axis be taken as the line joining q to the origin. ‘The dist b and the magnitde of ¢’ are to be determined in terms of the ep quantities: g, @, a. ‘The potential at an arbitrary point P due to q and q' is given by Ue, 8, 8) = On the surface of the sphere, r = a, and U(a, 6, @) = O for all 9 and ¢. But from expression (3-37), U(a, 6, 6} can equal zero for ali @ only if the two square roots are proportional to each other. This is the case if 6 = a?/d, for then Hence, (3-38) and furthermore, g=-50 (3-39) "These equations serve to specify the location and magnitude of the first image charge. A second image charge q’ may be placed et the center of the sphere without destroying the equipotential nature of the spherical surface. The magnitude of g” is arbitrary; it may be adjusted to fit the boundary conditions on the problem. Thus a complete solution to the point charge-311) LINE CHARGES AND LINE IMAGES 61 conducting sphere problem has been effected; the potential at all points exterior to the sphere is gi¢ U(r, 6, 6) = es tat (3-40) ‘The potential of the spherical conductor itself is 2 oe U(@, 6) = Fea (41) and the surface density of charge on the sphere is au ' (0, 6) = e055] (3-42) |e’ All the lines of force which would normally converge on the image charges are intercepted by the sphere; hence the total charge on the sphere, Q, is equal to the sum of the image charges: Q=a 4". co) This result may be verified by direct integration of (3-42). Special cases of interest are the grounded sphere: U(a) = 0, = and the uncharged spherical conductor: g’ = —q. 3-11 Line charges and line images. Thus far, our image technique has been limited to problems involving point charges, and hence point images. In this section we shall take up several problems which may be solved by means of line image charges. Consider two infinitely long, parallel, line charges, with charges \ and —A per unit length, respectively, as shown in Fig. 3-6. The aa at any point is given by - eon (344) Pla, y) a Bquipoteniis: “Se surface IT __ Equipotential surface I Fig. 3-6. Two infinitely long, parallel line charges (of charge \ and —A per unit length) are shown cutting the plane of the paper.62 SOLUTION OF ELECTROSTATIC PROBLEMS [cHap. 3 where 7; and rz are the perpendicular distances from the point to the two line charges. The equipotentials are obtained by setting (3-44) equal to a constant, a procedure which is equivalent to requiring that where Af is constant. Hence the équipotentials may be specified by (3-45). ‘The equipotential corresponding to M = 1 is the plane located halfway between, the two line charges, shown as equipotential surface I in the figure. ‘The potential of the plane is zero. Hence the problem of a long line charge oriented parallel to a conducting plane has been effectively solved. ‘The potential in the half-space is given correctly by (3-44). Let us assume that the line charge shown on the right side of the figure is the specified charge, which is at a distance d from the conducting plane. Then the line charge on the left side of the figure plays the role of an image. Again, the totai charge on the plane is equal to that of the image charge. Let us next consider equipotential surfaces corresponding to other values of M. The general form of the surface may be found by expressing ry and ra in rectangular coordinates. For convenience, we choose the origin of the coordinate system on the positive line charge, and make this charge coincide with the z-axis; we let the sccond line charge be located at a= —2d,y = 0. Now 2 ee Ze ory and rhe @+2dy+y, so that (3-45) becomes, after a little algebraic manipulation, are 1— M2? Ce 1a = ety (3-46) This is the equation of a circular cylinder extending paraliel to the z-axis. If Af is less than one, the cylinder surrounds the positive line charge, as does equipotential surface IT of the figure. The axis of the cylinder goes through the point (8-48) We are now in a position to solve a number of interesting problems involving cylindrical conductors, but only one of this type will be dis- cussed. Consider the problem of a long cylindrical conductor in the3-12] SYSTEM OF CONDUCTORS. COEFFICIENTS OF POTENTIAL 63 vieinity of a conducting plane, and oriented parallel to it. The cylinder bears the charge d per unit length. Figure 3-6 may serve to illustrate the problem, the two conductors coinciding with the dotted surfaces. Both of the line charges are images in this case, and the potential in the region surrounding the cylinder and to the right of the plane is given by (3-44). It is evident that the charge induced on the plane is equel.to —X per unit distance in the z-direction. 3-12 System of conductors. Coefficients of potential. In the preceding sections several important methods for obtaining solutions to Laplace’s equation have been discussed. Although general in scope, these methods are limited by practical considerations to problems in which the con- ductors have rather simple shapes. When their shapes are complicated, complete mathematical solution is out of the question; nevertheless, cer- tain conclusions can be drawn about the system just because the potential satisfies Laplace’s equation. In fact, we shall prove here that a linear relationship exists between the potential of one of the conductors and the charges on the various conductors in the system. The coefficients in this relationship, the so-called coefficients of potential, arc functions only of the geometry and, although not always calculable, may be determined directly from experiment. Suppose there are N conductors in fixed geometry. Let all of the con- ductors be uncharged except conductor j, which bears the charge Qj. The appropriate solution to Laplace’s equation in the space exterior to the conductors will be given the symbol U“(s, y, 2), and the potential of each of the conductors will be indicated by U1, U2, ...,Uj®,..., Un. Now let us change the charge of the jth conductor fo \Qj. The function dU“ (z, y, 2) satisfies Laplace’s equation, since A is a constant; that the new boundary conditions are satisfied by this function may be seen from the following argument. The potential at all points in space is multiplied by A; thus all derivatives (aud in particular the gradient) of the potential are multiplied by A. Because o = eo, it follows that all charge densities are multiplied by X. Thus the charge of the jth conductor is AQjo and all other conductors remain uncharged. A solution of Laplace’s equation which fits a particular set of bound- ary conditions is unique; therefore we have found the correct solution, AU (a, y, 2), to our modified problem. The interesting conclusion we draw from this discussion is that the potential of each conductor is pro- portional to the charge Q; of conductor j, that is, uy = pi, @=1,2,.. (3-49) where pj; is a constant which depends only on the geometry.64 SOLUTION OF ELECTROSTATIC PROBLEMS lenar, 8 The same argument may be applied to the case where conductor & is charged: O, = vQzo, all other conductors being uncharged. Here the appropriate sohition to Laplace’s equation is vU(z, y, 2), where U® is the solution for vy = 1. It is apparent, then, thet @, y, 2) + vO x, y, 2) (3-50) ue is a solution appropriate to the case where both conductors are charged. Again we appeal to the uniqueness of a solution for a given set, of boundary conditions. Thus (3-30) is the solution for this case, and the potential of each conductor may be written as Us = piQs + paQi, (= 1,2,...,N). (3-31) This result. may be generalized immediately to the case where all N conductors are chatged: (3-52) This is the linear relationship between potential and charge which we have been seeking; the coeflicients p,; are called the coefficients of potential. In Chapter 6 it will be shown that the array of these coefficients is sym- metrical, i.e., that p:; 3-13 Solutions of Poisson’s equation. In the preceding sections, we have dealt exclusively with Laplace’s equation and its solution. Laplace’s equation is applicable to those electrostatic problems in which all the charge resides on surfaces of conductors or is concentrated in the form of point or line charges. We shall see in the next chapter that it is necessary for only the free charge (i.e., the charge which is free to move or to be trans- ferred from one object to another) to be distributed in this manner; if the region between the conductors is filled with one or more simple dielee- tric media, then Laplace’s equation still holds in these media. Let us consider, now, an electrostatic problem in which part of the charge (the prescribed charge) is given by p(x, y, 2), a known function, and the rest of the charge (the induced charge) resides on the surfaces of eonductors. Such a problem requires the solution of Poisson’s equation. The general solution to this problem may be written 2s an integral of the type (3-1) over the prescribed charge plus a general solution to Laplace’s equation. ‘The solution to Laplace’s equation must be chosen, however, so that the entire potential satisfies all boandary conditions. When all of the charge is prescribed, i.e., when dg = p(w, y, 2) dv is known at all points.in space, then Eq. (3-1) represents the entire solution to Poisson’s equation, and this integral may be performed (either ana- lytically or numerically). There is one case, however, where the solationREFERENCES: 65 to Poisson’s equation may be obteined more direcly than by means of the formal solution (3-1); this occurs when both p and U are functions of only one independent variable. As an example of this case, let p be a function of the spherical coordinate, r, only, and let the entire charge be distributed in a sphericatly syrametrie way. Then (3-5b) becomes A & (# a) a Lon. (3-53) We shall assume that the total charge is bounded, ie., that either the charge doos not extend to invinity ov the charge density drops off sufficiently rapidly at large radii, Equation (3-53) may then be integrated directly, assuming the function p(ri given, and the two constants of integration may be determined (1) froin Gauss’ Law for the eleétrie field at some ‘The following texts ave recommended for (1) a more complete discussion of Legendre’s equation, (2) the general form of Laplace’s equation in orthogonal, curvilinear coordinates, and (3) a more complete discussion of the solution to Laplace’s equation: H. Maroewau and G.M Munpiy, The Mathematics of Physics and Chemistry, D. Van Nostrand Co., Juc., New York, 1943. J. A. Srravton, Electromagnetic Theory, McGraw-Hill Book Co., Ine., New ‘York, 1943. W. Panorsxy and M. Pratuses, Classical Electricity and Magnetism, Addison- Wesley, 1955.66 SOLUTION OF ELECTROSTA‘tiy PROBLEMS [cnap. 3 PRopems 3-1. Two spherical conducting shells of radii r, and ry a: centrically and are charged to the potentials U, and Us, respecti find the potential at points between the shells, and at points r > s 3-2. Two long cylindrical shells of radii re and ry are arranged coaxially and are charged to the potentials U, and Us, respectively. Find the potential at points between the cylindrical shells. 3-3. If U1 is a solution to Laplace’s equation, prove that the partial derive~ tive of U1 with respect to one or more of the rectangular coordinates (e.g., 9U;1/dx, 8?U1/d2?, 8°U 1/axdy, etc.) is also a solution. 3-4. Show that half the zonal harmonics are generated by differentiating r~! successively with respect to the rectangular coordinate z (2 = r cos 6). 3-5. Obtain V?U in cylindrical coordinates (Eq. 3-8), from the rectangular form, (3-6), by direct substitution: z = r cos @, y = rsin 8. 3-6. Find the potential of an axial quadrupole: point charges ¢, —2g, q placed on the zaxis at distances J, 0, —I from the origin. Find the potential oniy at distances r >> 1, and show that this potential is proportional to one of the zonal harmonics. 3-7. A conducting sphere of radius @ bearing total charge Q is placed in an initially uniform electric field Eo. Find the potential at: ail points exterior to the sphere. 3-8. A long cylindrical conductor of radius a bearing no net charge is placed in an initially uniform electric field Eo. The direction of Eo is perpendicular to the cylinder axis, Find the potential at: points exterior to the cylinder, and find also the charge density on the cylindrical surface. *3-9. Show that Im A{(z-+ jy)? = Ar!’ sin 40 satisfies Laplace’s equa- tion, but that the electric field derived ee this function has a discontinuity at @ = 0. (Note that r and @ are cylindrical coordinates here.) The function may be used to describe the potential at the edge of a charged conducting plane. ‘The conducting plane coincides with the rz-planc, but only for positive values of x, Find the charge density on the plane. Make a sketch showing several equipotential surfaces and several lines of force. 8-10. A point charge 9 is situated a distance d from a grounded conducting plane of infinite extent. Obtain the total charge induced on the plane by direct integration of the surface charge density. 3-11, Two point charges, q: and qz, are located near a conducting plane of infinite extent. Find the image charges which are needed to make the plane a surface of constant potential. From the result just obtained, can you predict the image charge distribution required for the case of a body of arbitrary shape with charge density ¢ situated near a conducting plane of infinite extent? 3-12. Find the force between a point charge q and an uncharged conducting sphere of radius a. The point charge is located a distance r from the center of the sphere, where r > a. arranged con: . ity > tay * Starred problems are more difficult.PROBLEMS: 87 3-18. Show that the problem of an uncharged conducting sphere in an initially uniform electric field Eo may he solved by means of images. [Hint: A uniform clectric field in the vicinity of the origin may be approximated by the field of two point charges Q and —Q placed on the eaxis at 2 = —L and z = -+L, respectively. The field becomes more nearly uniform as L — ©. Tt is evident that Q/2reoL? = Eo] 3-14. A point charge 9 is located inside and at distan spherical conducting shell. ‘The inner radius of the sh problem can be solved by the image technique, and find the charge density + induced on the inside surface of the shell. (The potential of the spherical shell cannot be completely specified in terms of g and ite image, bécause exterior fixed charges can also contribute. Nevertheless, these exterior charges will add only & constant term to the potential.) Find the total charge induced on the inside surface of the shell (a) by physical arguments, and (b) by integration of o over the surface. 3-15. A long conducting cylinder besring a charge ) per unit length is oriented paralicl to a grounded conducting plane of infinite extent. The axis of the linder is at distance 29 from the plane, and the radius of the eyiiuder is 4. Find the location of the line image, and find also the constant M {which de- termines the potential of the cylinder) in terms of @ and zo. 3-16. A spherical distribution of charge is charecterized by a c density p fors < R. For radii greater than R, the charge den: the potential U(r) by integrating Poisson's equation. ovaluating the integral (3-1). [Hint: To perform (3-1), divide the charge region into spherical concentric shells of thickness dr.] 3-17. A dipele p is oriented normal to and at distance d fro coaducting plane. The planc force exerted on the plane by the dipole. 3-18. A thunderstorm contains a charge -+@ at altitude Ai aad, directly below this, a charge —Q at altitude he. Find an expression for th y cicetric ficld By at the earth’s surface at distance d from the storm. Por hi = 5000m, ha = 3000m, and Q = 15 coul, make s graph showing how Ey varies, from d = 0 tod = 20km. from the center of a how that this stant char, an infin s grounded (i.e., at zero potential). Calculate theCHAPTER 4 THE ELECTROSTATIC FIELD IN DIELECTRIC MEDIA Thus far, we have ignored problems involving dielectric media, and have dealt with cases in which the electric field is produced exclusively by free charges: either by a specified distribution of thera or by free charge on the surface of conductors. We now wish to remedy this situation and take up the more general ease An ideal dielectric material is one which has no free charges. Neverthe- less, all material media ste composed of molecules, these in turn being composed of charged entities {atomic nuclei and electrons), and the molecules of the dielectric are certs! affected by the presence of an elecivie field. The electric field causes a force to be exerted on each charged particle, positive particles being pushed in the direction of the field, nega- tive particles oppositely, sc that the positive and negative parts of each molecule are displaced from their equilibrium positions in opposite direc- tions. These displacements, however, are limited (in most cases to very small fr: s of a molecular diameter) by strong restoring forces which ave set up by the changing charge configuration in the molecule. The over- all effect from the macroscopic point of view is most easily visualized eut of the entire positive charge in the dielectric relative to the negative charge. The dielectric is said to be polarized. A polarized dielectric, even though it is electrically neutral on the average, produces an electric field, both at exterior points and inside the dielectric as well. As a result, we are confronted with what appears to be an awkward tion: the polarization of the dielectric depends on the total electric fieid in the medium, but a part of the electric field is produced by dielectric itself. Furthermore, the distant electric field of the dielectric may modify the free charge distribution on conducting bodies, and this in turn will change the electric field in the dielectric. It is the main purpose of this chapter to develop general methods for handling this curious situation. 4-1 Polarization. Consider a small volume element Av of a dielectric medium which, as a whole, is electrically neutral. If the medium is polar- ized, then a separation of positive and negative charge has been effected, and the volume element is characterized by an electric dipole moment Ap = fra (4-1) According to Section 2-9, this quantity determines the electric field ” 68+1) POLARIZATION 69 produced by Av at distant points (i.e.; at distances from Av large com- pared with the dimensions of the volume element). Since Ap depends on the size of the volume element, it is more con- yenient to work with P, the electric dipole moment per unit volume: By 7 (4-2) Strictly speaking, P must be defined as the limit of this quantity as Av becomes very small from the macroscopic viewpoint. In this way P becomes a point function, P(x, y,z). P is usually called the electric polar- ization, or simply the polarization, of the medium, Its dimensions are charge per unit area; in mks units, coul/m?. It is apparent that P(x, y, z) is a vector quantity which, in each volume element, has the direction of Ap. This, in turn, has the direction of dis- placement of positive charge relative to negative charge (sec Fig. 4-1). Fic, 4-1. A piece of polarized dielectric material. Each volume element is represented as a dipole Ap. Although Ay is assumed very small from the macroscopic point of view, it still contains many molecules. It is sometimes desirable to speak about the electric dipole moment of a single molecule, that is, Pm = rdq, (4-3) molecule since a molecule is one of the small, electrically neutral entities which make up the dielectric material. It is evident from (4-1) that the dipole moment associated wigh Av is given by Ap = Spm, where the summation extends over all molecules inside the element Av. Hence, P= Dw (4) This approach will be developed further in Chapter 5. Although Fig. 4-1 represents each volume element of the polarized dielectric as a small dipole, it may be more instructive to visualize the dielectric in terms of its molecules, and to imagine that each dipole of Fig. 4-1 represents a single molecule.70 ELECTROSTATIC FIELD IN DIELECTRIC MEDIA. [cuar. 4 4-2 External field of a dielectric medium. -Consider now a finite piece of dielectric material which is polarized, i.e., which is characterized at each point r’ by a polarization, P(r’). The polarization gives rise to an electric field, and our problem is to calculate this field at point r, which is outside of the dielectric body- (see Fig. 4-2). As in Chapter 2, we shall find it more convenient to calculate first the potential U(r), and obtain the electric field as minus the gradient of U. Each volume element Av’ of the dielectric medium is characterized by a dipole moment Ap = P Av’, and since the distance between (z, y, 2) and Av’ is large compared with the dimensions of Av’, this quantity (the dipole moment) completely determines Avs contribution to the potential: Y) _ P@’):@—r’) av’ aoe ——rs—OS™e—O (4-5) AU®) = Here r — 1’ is the vector, directed cut from Av’, whose magnitade is given by eor/=VEeP TOPE 8) The entire potential at point r is obtained by summing the contributions from all parts of the dielectric: _ 1. Pe) ra be U@ = ge , ropa (7) ‘This result is correct, and U may be evaluated directly from (4-7) if the functional form of P is known. It will be to our advantage, however, to express (4-7) in a rather different way by means of a simple mathematical transformation. . @ ya) cf Fic. 4-2. The electric field at (x, y, z) may be calculated by summing up the contributions due to the various volume elements AV’ in Vo. The surface of Vo is denoted by So.42] EXTERNAL FIELD OF A DIELECTRIC MEDIUM 7 If |r — ¥'| is given by (4-6), then a 1 _ ror w (ta) =+ pote (4-8) as may be seen by direct application of the gradient operator in cartesian coordinates. The V’ operator involves derivatives with respect to the primed coordinates. In certain circumstances it may be desirable to per- form a gradient operation with respect to the unprimed coordinates; this will be indicated in the usual way by V. Evidently, V’ operating on a function of |r — r’| is equal to —V operating on the same function. We shall require the V operator later in order to get the electric field at point r. However, in performing the integral (4-7) over the dielectric volume Vo, the point r is held fixed; hence the integrand of (4-7) may be trans- formed by means of (4-8): Pita). py (4 ). (5) F-?R FFI Equation (4-9) may be further transformed by means of the vector identity (I-6) of Table 1-1: div’ (fA) = fdiv’ A+ A-V‘f, (4-10) where f is any scalar point function and A is an arbitrary vector point function. Here again the prime indicates differentiation with respect to the primed coordinates. Letting f = (1/|r — 1’|) and A = P, the inte- grand, (4-9), becomes EG) Finally, the potential, (4-7), may be written as I ee eda oe fC diveP ar : UO) = Fre £2 : 4meyIsyrs' [FF] (423) where Vo — V1 is the volume of the dielectric excluding the “needle,” So ig the exterior surface of the dielectric, and S’ = S; -+ Sz + S, are the needle surfaces. But from Fig. 4-3 it is seen that op = 0 on the cylindrical surface 8, of the needle; furthermore, the needle may be made arbitrarily thin so that the surfaces S; and Sp have negligible area. Thus only the exterior surfaces of the dielectric contribute, and the surface integral of Eq. (4-23) becomes identical in form to the surface integral of Eq. (4-15). The volume integral of Eq. (4-23) excludes the cavity ; however, the contribution of the cavity to this integral is negligible, as may readily be seen. The charge density pp is bounded; the quantity dv'/|r — r'j does not diverge at the field point (i.e., when x’ = 1) because the volume of a point is a higher-order zero than the lim |r — r/|; and finally volume V, of the needle may be made arbitrarily small by making the cavity thin, Thus we need not exclude the volume Vj, and Eq. (4-23) becomes similar in form to Eq. (4-15). In other words, Eq. (4-15) gives the potential U(x) regardless of whether the point r is Jocated inside or external to the dielectric. ‘The electric field E(x) may be calculated as minus the gradient of Eq. (4-23). But this differs only by a negligible amount from Eq. (4-21). Thus (4-21) gives the medium’s contribution to the electric field at r, inde- pendently of whether x is inside or outside the medium. Tre calculations indicated in Eqs. (4-15) and (4-21) are straight- forward for cases in which P(e, y, z) is a known function of position, (Some examples of this type are to be found among the problems at the end of this chapter.) In most cases, however, the polarization arises in re- sponse to an electric field which has been imposed on the dielectric medium [that is, P(x’, y’, 2’) is a function of the total macroscopic electric field E(2’,y’,2’)|, ond under these conditions the situation is much more complicated. Hirst, it is necessary to know the functional form of P(E); but this is known experimentally in most cases and hence is not a source of difficulty, ‘The real complication arises because P depends on the total electric field, including the contribution from the dielectric itself, and it ig this contribution which we are in the process of evaluating. Thus we cannot determine P because we don’t know E, and vice versa. It is evident that a d:fferent approach to the problem is needed, and this will be provided in thy following sections.4-4] GAUSS’ LAW IN A DIELECTRIC 77 4-4 Gauss’ law in a dielectric. The electric displacement. In Chapter 2 we derived an important relationship between electric flux and charge, namely, Gauss’ law. This law states that the electric flux across an arbi- trary closed surface is proportional to the total charge enclosed by the surface, In applying Gauss’ law to a region containing free charges em- bedded in a dielectric, we must be careful to inelude all of the charge in- side the gaussian surface, bound charge as well as free charge. ‘Dielectric medium Fig. 4-4. Construction of a gaussian surface S in a dielectric medium. In Fig. 4~4 the dashed surface $ is an imaginary closed surface located inside a dielectric medium. There is a certain amount of free charge, Q, in the volume bounded by S, and we shall assume that this free charge exists on the surfaces of three conductors in amounts q1, go, and qj. BY Gauss’ law, 1 Bn aa = L@+an, a2 where Q is the total free charge, @ and Qp is the polarization charge: 1 Qa + da, Qp = Leer P-nda+ f, (div P) de. (4-252) Here V is volume of the dielectric enclosed by S. There is no boundary of the dielectric at §, so that the surface integral in (4-25a) does not contain a contribution from S. If we transform the volume integral in (4-25a) to.a surface integral by means of the divergence theorem, we must be careful to include contribu- tions from all surfaces bounding V, namely, S, $1, So, and Ss. It is evident that the last three contributions will cancel the first. term of (4-25a), so that Qp = — $, P-nda, (4-25b)78 ELUCPROSPATIC FIELD IN DIELECTRIC MEDTA [onar. 4 Combining this result with (4-24), we obtain ¢, (6B + P)-nda = Q. (4-26) Equation (4-26) states that the flux of the vector é9E + P through a closed surface is equal to the total free charge enclosed by the surface. This vector quantity is important enough to warrant a name and a separate symbol. We define, therefore a, new macroscopic field vector D, the electric displacement: D= &E+P, (4-27) which evidently has the same units as P, charge per unit area. In terms of D, Eq. (4-26) becomes fd nda=Q (4-28) and this result is usually refer:ed to as Gauss’ law for the electric displace- ment, or simply Gauss’ law. Equation (4-28) is applicable to a region of space bounded by any closed surface S; if we apply it to a smail region in which all of the free charge enclosed is distributed as a charge density p, then Gauss’ law becomes $, D-nda = pav. Dividing this equation by AV and proceeding to the limit, we obtain divD = p, (4-29) a result which is sometimes called the differential form of Gauss’ law. The advantage of the procedure just followed is that the total electro- static field at each point in the dielectric medium is expressed as the sum of two parts, E(e, y, 2) = Ere, ne — EPG, n 2, (4-30) where the first term, (1/€0)D, is related to the free charge density through its divergence, and the second term, (—1/€o)P, is proportional to the polarization of the medium. In a vacuum the electric field is given entirely by the first term in (4-30). 4-5 Electric susceptibility and dielectric constant. In the introduction to this chapter it was stated that the polarization of a dielectric medium occurs in response to the electric field in the medium. The degree of polarization depends not only on the electric field, but also on the proper+ ties of the molecules which make up the dielectric material. From the!4-5] ELECTRIC SUSCEPTIBILITY AND DIELECTRIC CONSTANT 79 macroscopic point of view, the behavior of the material is completely specified by an experimentally determined relationship, P = P(E), where E is the macroscopic electric field. ‘This is a point relationship, and if E varies from: point to point in the material, then P will vary accordingly. For most materials, P vanishes when E vanishes. Since this is the usual behavior, we shall limit our discussion here to materials of this type. (Dielectrics with a permanent polarization will be discussed briefly in Section 5-4.) Furthermore, if the material is isotropic, then the polariza- tion should have the same direction as the electric field which is causing it. These results are summarized by the equation P= x(Z)E, (4-31) where the scalar quantity x(Z) is called the eleciric susceptibility of the material. A great many materials are electrically isotropic; this category includes fluids, polycrystalline and amorphous solids, and some crystals. ‘A treatment of the electrical properties of anisotropic materials is beyond the scope of this text. Combining (4-31) with (4-27), we obtain an expression for D in iso- tropic media: D = «(Z)E, (4-32) eZ) = &) + x(B), (4-33) where e(K) is the permittivity of the material. It is evident that ¢, €, and x all have the same units. ‘Although we have been careful to write x and ¢ in the form x(Z) and ¢(E), nevertheless it is found experimentally that x and ¢ are frequently independent of the electric field, except perhaps for very intense fields. In other words, X and ¢ are constants characteristic of the material. Ma- terials of this type will be called linear dielectrics, and they obey the relations P=%xs, (4-81a) D=«. (4-32a) ‘The electrical behavior of a material is now,completely specified by either the permittivity ¢ or the susceptibility x. It is more convenient, however, , to work with a dimensionless quantity K defined by e= Ke. (4-34) K is called the dielectric coefficient, or simply the dielectric constant. From (4-33) it is evident that K==-=1+ (4-35) Sia S\x80 ELECTROSTATIC FIELD IN DIELECTRIC MEDIA {cnar. 4 The dielectric constants for a few commonly encountered materials are given in Table 4-1. Except for a few examples in which the polarization P of the material is specified, the problems in this book deal with linear dielectrics. If the electric field in a dielectric is made very intense, it will begin to pull electrons completely out of the molecules, and the material will be- come conducting. The maximum electric field which a dielectrié can withstand without breakdown is called its dielectric strength. The dielec- tric strengths Emax, of a few substances are also given in Table 4-1. TaBLe 4-1 Properties or Dretectric MaTerrats* (Dielectric constant K and diel ctric strength Emax) i ; Material | Exax, Volts/m | Glasst 9x 10° Mica 5-20 X 10° "Nylon 16 X 10° Rubbert * 16-40 x 108 | | Sulfur | Woodt Alcohol, ethyl (0°C) Benzene (0°C) | Petroleum oil 12 x 10° Water (distilled, 0°C) Water (distilled, 20°C) Air (1 atm) 3x 10% Air (100 atm) COz (1 atm) * Data from the Handbook of Chemistry and Physics, 33rd edition, Chemical Rubber Publishing Co., Cleveland, Ohio. { For materials such as glass, rubber, and 3wood, the chemical composition varies; hence the range of dielectric constants. It is not to be inferred that the material is nonlinear. 4-6 Point charge in a dielectric fluid. One of the simplest preblems involving a dielectric which we might consider is that of a point charge q in a homogeneous isotropic medium of infinite extent. The dielectri¢ medium will be assumed to he linear and characterized by a dielectric4-6] POINT CHARGE IN A DIELECTRIC FLUID 81 constant K. Although this problem is quite simple, it will nevertheless prove instructive. If the point charge q were situated in a vacuum, the electric field would be a pure radial field. But since E, D, and P are all parallel to one another at each point, the radial nature of the field is not changed by the presence of the medium. Furthermore, from the symmetry of the problem, E, D, and P can depend only on the distance from the point charge, not on any angular coordinate. Let us apply Gauss’ law, Eq. (4-28), to a spherical surface of radius r which is located concentrically about 9. For c venience, q will be located at the origin. Then 4nr?D = q und =, D= dor? or D= 7a r (4-36) (4-37) (4-38) Thus the clectric field is smaller by the factor K than would be the case if the medium were absent. At this point, it will be instructive to look at the problem in more detail, and try to see why the diclectric has weakened the electric field. The electric field has its origin in all of the charge, bound and free. The free charge is just the point charge g. The bound charge, however, is made up from two contributions, a volume density pp — -- div P, and a surface density ¢p = P-n on the surface of the diclectrie in contact with the point charge. Using Eq. (4-38), we find that div P vanishes, so there is no volume density of bound charge in this case. Our point charge ¢ is a point in the macroscopic sense. Actually, it is large on a molecular scale, and we can assign to it a radius b which even- tually will be made to approach zero. The total surface bound charge is then given by = ii 2(p. oo eae Qp = lim 4n67(P-m) ran = K The total charge, 1 Qe+g= Re (4-40)82 ELECTROSTATIC FIELD IN DIELECTRIC MEDIA [cnar. 4 Fic, 4-5. Schematic diagram showing the orientation of polarized molecules in a dielectric medium surrounding a “point charge” q. appears as a point charge from the macroscopic point of view, and it is now clear why the electric field is a factor K smaller than it would be if the medium were absent. A schematic diagram of the point charge q in a dielectric medium is shown in Fig. 45. 4-7 Boundary conditions on the field vectors. Before we can solve more complicated problems, we must know how the field vectors E and D change in passing an interface between two media. The two media may be two dielectrics with different properties, or a dielectric and a conductor. ‘Vacuum may be treated as a dielectric with permittivity eo. Censider two media, 1 and 2, in contact as shown in Fig. 4-0. We shall assume that there is a surface density of free charge, ¢, which may vary frorn point to point on the interface. Let us construct the small pillbox- shaped surface S which intersects the interface and encloses an area AS of the interface, the height of the pillbox being negligibly small in com- parison with the diameter of the bases. The free charge erclosed by S is o AS + (ex + ps) X volume, but the volume of the pillbox is negligibly small, so that the jast term may be neglected. Applying Gauss’ law to S, we find Dz-ny AS + D,-0, 48 = o AS, or (Bz — Di) - Be (4-41a) Since nz may serve as the normal to the interiace, Dan — Di (4-41b(£7] BOUNDARY CONDITIONS ON THE FIELD VECTORS 83 Fira, 4-6. Boundary conditions on the field vectors at the interface between two inedia may be obtained by applying Gauss’ law to S, and integrating E- dl around the path ABCDA, Thus, the discontinuity in the normal component of D is given by the surface density of free charge on the interface. Or, to put it another way, if there is nv free churge on the interface between tivo media, the normal component of D is continuous. Beeause the electrostatic field E may be ubtained as minus the gradient of a potential, the iine integral of Edi around any closed path vanishes. Let us apply this result to the rectangular path ABCD of Fig. 4-6. On this path, the lengths 4B and CD will be taken equal to Al and the seg- ments AD and BC will be assumed to be negligibly small. Therefore E,-Al +E, -(~al) = 0, or (4-42a) Hence, the desired result: Eo = Evy (4+42b) that is, the tangential component of the clectrie field is continuous across an interface. The above results have been obtained for two arbitrary media, but ia is worth our while to specialize the equations for the case where one of the media is a conductor. If medium 1 is taken as the conductor, then E, = 0. Since E, vanishes, there is no polarization, and by Eq. (4-27) D, also vanishes. Thus (4~41h) and (4-42b) become Don = 9, (4-48) (4-44) for the displacement and electric field in a dielectric just outside of a vondustor.84 ELECTROSTATIC FIELD TX DIELECTRIC MEDIA {omar. 4 It is evident. on purely physical grounds that the potential U must be continuous across an interface. This follows because the difference in potential, AU, between two closely spaced points is —E- Al, where Al is the separation of the two points, and from what has been said above there is no reason to expect E to become infinite at an interface. Actually, the continuity of the potential is a boundary condition, but not inde- pendent of those already derived. It is equivalent in most cases to (4-42b). From the discussion above and in preceding sections, it may be inferred that the electric displacement D is closely related to free charge. We should now like to prove an important property of D, namely, that the flux of D is continuous in regions containing no free charge. To do so, we aguin resort to Gauss’ law. Let us focus our attention on a region of space and construct lines of displacement, which are imaginary lines drawn in * such a way that the direction of a live at any point is the direction of D at that point. Next we imagine a tube of displacement, a volume hounded on the sides by lines of D but not cut by them (see Fig. 4-7). The tube is terminated at its ends by the surfaces S, and S. Applying Gauss’ law, we obtain D-nda — f, D-n'da = Q. (4-45) Sa Fig. 4-7. A tube of displacement flux. If there is no free charge in the region, then Q@ = 0, and the same amount of flux enters the tube through S, as leaves through Sp. When free charge is present, it determines the discontinuity in displacement flux; thus lines of displacement terminate on free charges. The lines of force, on the other hand, may terminate on free or bound charges. 4-8 Boundary-value problems involving dielectrics. The fundamental equation which has been developed in this chapter is divD = p, (4446)4-9) DIELECTRIC SPHERE IN A UNIFORM ELECTRIC FIELD 85 where p is the free charge density. If the dielectrics with which we are concerned are linear, isotropic, and homogeneous, then D = ¢E, where ¢ is a constant characteristic of the material, and we may write divE = i p. (4447) But the electrostatic field E is derivable from a scalar potential U, i.e., E = —grad U; so that VU = — dp. (4448) ‘Thus the potential in the dielectric satisfies Poisson’s equation; the only difference between (4-48) and the corresponding equation for the poten- tial in vacuum is that € replaces €. In most cases of interest the dielectric contains no free charge dis- tributed throughout its volume, that is, p = 0 inside the dielectric ma: terial. The free charge exists on the surfaces of conductors or is concen- trated in the form of point charges which may, to be sure, be embedded in the dielectric. In these circumstances, the potential satisfies Laplace’s equation throughout the body of the dielectric: vu =0. (49) In some problems there may be a surface density of free charge, , on the surface of a dielectric body or on the interface between two dielectric materials, but this does not alter the situation, and Eq. (4-49) still applies so long asp = 0. An electrostatic problem involving linear, isotropic, and homogeneous dielectrics reduces, therefore, to finding solutions of Laplace’s equation in each medium, and joining the solutions in the various media by means of the boundary conditions of the preceding section. ‘There are many problems which may be solved by this method; one example will be dis- cussed here and additional examples will be found in the problems at the end of the chapter. 4-9 Dielectric sphere in a uniform electric field. We should like to determine how the lines of force are modified when a dielectric sphere of radius a is placed in @ region of space containing an initially uniform elec- trie field, Eo. Let us assume the dielectric to be linear, isotropic, and homogeneous, and to be characterized by the dielectric constant K. Furthermore, it bears no free charge. The origin of our coordinate system may be taken at the center of the sphere, and the direction of Eo as the86 ELECTROSTATIC FIELD IN DIELECTRIC MEDIA [onar. 4 polar direction (z-direction); the potential may then be expressed as a sum of zonal harmonics. Just as in Section 3-5, all boundary conditions can be satisfied by means of the two lowest-order harmonics, and we write Uy(r, 8) = Ayr cos @ + Cyr? cos 6 (4-50) for the vacuum region (1) outside the sphere, and Ual(r, 6) = Aor cos 6 + Cor? cos 8 (4-51) for the dielectric region (2). The constants A1, A, Ci, and C’z are unknown and must be determined from the boundary conditions. The harmonic r7 is not required, since its presence implies a net charge on the sphere. A constant term may be added to (4-50) and to (4-51), but since the eame constant is required in both equations, we may, without loss of generality, take it to be zero. At distances far from the sphere, the electric field will retain its uniform character, and U;—» —Eor cos. Hence’ A; = —Eo. Furthermore, unless C2 = 0, the potential and associated electric field would become infinite at the center of the sphere, and this would imply the existence of a macroscopic dipole at the center, i.e., a dipole whose moment is not proportional to AV. Certainly, this is not the case; as has already been discussed in Section 4-3, the potential and macroscopic field do not become infinite in a dielectric devoid of free charges. Hence C'z = 0, and the remaining constants, Az and C;, may be obtained from the boundary conditions of Section 4-7. Continuity of the potential across the interface between the dielectric and vacuum requires that U; = U2 at r = a, or —Eoa + Cya~? = Aga. (4-52) Since the normal component of D at the interface is D, = —e(@U/ér), the continuity of D, (there is no free charge on the surface of the dielectric) requires that Di, = Day at r = a, or Eo + 2C\a~3 = —KAp. (4-53) Continuity of Z, at r = a is equivalent to (4-52). Combining Eqs. (4-52) and (4-53), we obtain ee eSBoe 4 Ag K+ (4-54) and n (K — 1)a*hy o="es: (4-85)4-10] FORCE ON A POINT CHARGE EMBEDDED IN A DIELECTRIC 87 (a) (b) Fic, 4-8. A uniform electric field is distorted by the presence of a dielectric sphere: (a) lines of electric displacement, (b) lines of the electric field. Thus the problem has been solved. The potential is given by (4-50) or (4-51), and the constants Ay, C1, A», and C3 are all known. The com- ponents of E and D may be obtained at any point (r, 8, 4) by differentia tion. It is evident from (4-54) and because C'z = 0 that the electric field inside the sphere has the direction of Ey and is given by The lines of displacement and the lines of force are shown in Fig. 4-8. 4-10 Force on a point charge embedded in a dielectric. We are now in a position to determine the force on a small, spherical, charged conductor embedded in 2 linear, isotropic dielectric. In the limit in which the con- ductor is negligibly small from the macroscopic viewpoint, this calculation gives the force on a point charge. ‘The electric field and surface charge density at a representative point of the conductor surface will be obtained by the boundary-value pro- cedure of the preceding section, and the force F may then be obtained from the integral over the surface: Fe 4, B’o da. (4-87) Here E’ stands for the electric field at the surface element da minus that part of the field produced by the element itself. In other words, = E-—E, (4-58) where E, is the electric field produced by the surface element of charge, oda. It is important that E, not be included in the field E’, because the88 ELECTROSTATIC FIBLD IN DIRLECTRIC MBOTA fonar. 4 quantity E,o da represents the interaction of the charge element o da with its own field; this self-interaction clearly produces no net force on the element, but gives rise to a surface stress = oF, (4-59) ich is due to the mutual repulsion of the electrons (or of the excess positive ions) in the surface layer. This stress is balanced by strong co- hesive forces in the material of which the element is composed. It should be pointed out that when calculating forces on charged objects in Chapters 2 and 3, we implicitly subtracted the self field E,; thus, when calculating the force on a point charge, the field produced by the point charge was not included. A further discussion of the forces on charged objects will be taken up in Section 6-8. It may appear that the self field of the charged surface element o da is negligible because the element is of infinitesimal size. This, however, is not the case. The element is smail from the macroscopic point of view, to be sure, but one never quite goes to the limit. At a point directly on its surface, the clement appears to be an infinite plane, i.c., the clement subtends an angle of 27; hence, where n is a normal to the element and ¢ is the permittivity of the dicle: in contact with it. Thus the stress $, is proportional to o”, and is always a tension regardless of the sign of o. It is our purpose here to calculate the force on a conductor. Using the boundary conditions of Section 4-7, the total electric field at the con- ductor is given by (4-61) Combining (4-58), 44-80), und (4-63), we obtain and the force on the conductor becomes F=3 ¢ Eo da. (4-572) s Let us now fix our attention on @ small spherical conductor embedded in a dielectric of infinite extent. ‘The total charge on the conductor is Q; its radius is a. Since we shall eventually go to the limit in which a becomes very small, and since variations in the electric field (if they exist) are on a macroscopic scale, it is sufficient to consider the case in which the electric4-10] FORCE ON A POINT CHARGE EMBEDDED IN A DIELECTRIC 89 field is initially uniform in the neighborhood of the conductor. Let us denote this uniform field by the symbol Ey. The picture is similar to that of the boundary-value problem we solved in Section 3-5, except that here the conducting sphere is embedded (or immersed) in a dielectric of per- mittivity ¢, and in addition bears a net charge Q. By analogy with Section 3-5 we easily determine: the potential, Ur, 8) = Uo — Bor cos 0 + #2 cos 9+ 2; (4-62) 2 ° 2 Srer’ the electric field, E, = Eo(1 + 2a°/r°) cos 0 + Q/4mrer?, (4-63) Ey = —Eo(1 — a®/r5) sin 6; and the surface charge density on the surface of the sphere, 00). = ely! | = 3eHig cos @ + Q/4mra?, (4-64) The force may now be determined from Eq. (4-57a). By symmetry, the only nonzero component of force is that in the direction @ = 0, i.e., in the z-direction: Fi=4 if ” (Ex)rma 008 86(6)22r0? sin 8 46 = EQ, (4-65a) or F = QEp. (4-65b) This result is unchanged as we go to the limit of small a. Thus the electric field in the dielectric, Eo, is in agreement with the fundamental definition, namely, the force on a small test charge Q divided by the magnitude of Q.IN DISLECTRIC MEDIA fear. 4 ELECTROSTATIC FI ProsLEms 4-1. A thin dielectric rod ross section A extends along the z-axis from x = Gtow = L, The polarisation of the rod is dong ifs length, and is given by P, = az*+b, Find the volume density of polarization charge and the surface polarization charge on ench end. Show explicitly that the total bound charge vanishes in this case. 4-2. A dielectric cube of side L has # racial polarisation given by P = Ar, where A is a constant, and r = iz + jy + ke. ‘The origin of coordinates is at the center of the cube. Find all bound charge densities, and show explicitly that the total bound charge vanishes 4-3, A dieiectric rod in the shape of a right virewlar cylinder of length L and radius 2 is polarized in the direction of its length. If the polarisation is uniform and of magnitude P, caleaiate the electric field resulling from this polarization at 9 point on the axis of the rod. 4-4. Prove the following relationship between. the polarization, P, and the bound charge densities pp and op, for a dielectric specimen of volume V and surface 8. [Pee i, por do + i opr da, Hore, r = it + jy-+ ke is the position vector from any fixed origin, (Hint: Expand div @P) according to Eq. (4~10)J 4-6. Two semi-infinite blocks of dielectrit are placed almost in contact so that there exists » narrow gap of constant separation between them. The polarization P is constant throvghout all of the diolectrie material, and it makes the angle 7 with the normal to the planes bounding the gop. Determine the electric field in the gap. 4-6, A long cylindrical conductor of railius a, bearing the charge A per unit length, is immersed in a chelectric medium of constant permittivity « Find electric field at distance 7 > a from the axis of the eylinder. 4-7, Two dielectric media with dielectric consinats Ky and K2 are separated by aplane interface. There is no free charge on the interface. Find a relationship between the angles 6; aud 62, where these are the angles which an arbitrary line of displacement makes with the norma! to the interface: 6: in medium 1, 42 in medium 2. 4-8. A coaxial cable of circular cross section has a compound dieleetric, The inner conductor has an outside radius a; this is surrounded by a dielectric sheath of dielectric constant Ky and of outer radius 6. Next comes another dielectric sheath of dielectric constant Ky and outer vadius c. {fs potential difference Vo is imposed between the conductors, caleulate thy polarization at each point in the two dictectric media. *4-9. Two dielectric media with countant permittivities «1 and ep ar cated by « plane interince. Phere is uo 2 ou the interface. A. point charge q is embedded in the medium characterized by ¢;, at s distance d from the inter- face. For convenience, we may take the ye-plane through the origin to be the interface, and we iocate 7 on the a-axis at z = —d. IfPROBLEMS 91 r=VerdgyPte, and rf = Ved? FPF], then it is easily demonstrated that (1/4zre1){(q/r) -+ (q’/r’)] satisfies Laplace’s equation at all points in medium 1 except at the position of g. Furthermore, q/4xezr satisfies Laplace’s equation in medium 2. Show that all boundary conditions can be satisfied by these potentials, and in so doing determine 4’ and q. (Refer to Fig. 4-9.) LLL GZ ® Ly Fioure 4-9 4-10. A long diclectric cylinder of radius a and diclecteie vonstart & iv placed in a uniform electric ficld Eo. The axis of the cylinder is oriented at right angles to the direction of Eo. The cylinder contains no free charge. Determine the electric field at points inside and outside the cylinder. 4-11. Two parallel conducting plates arc separated by the distance d and maintained at the potential difference AU. A dielectric slab, of dieleetria con- stant K and of uniform thickness ¢ < d, is inserted between the plates. De- termine the field vectors E and D in the dielectric and alse in the vacuum between dielectric and one plate. Neglect edge effects due to the finite sizo of the plates. 4-12, Two parallel conductivg plates ure separated by the distance d and maintained at the potential difference AU. A dielectric slab, of dielectric con- stant K and of aniform thickness d, i d sougly between the plat however, the slab does not comple the plates. Find the electric field (a) in the dielectric, and /) in the vacuum region between the plates. Find the charge density o on that part of the plate (c) in contact with the dielectric, and (d) in contact with vacuum. (e) Tind ep on the surface of the dielectric slab. ; 4-18, A conducting sphere of radius R floats half submerged in a liquid Jiclee- tric medium of permittivity ¢;. The region above the liquid is a gas of permit- tivity ee. The total free charge on the sphere is Q. Wind @ radial inverse-square field satistying all boundary conditions, and determine the free, bound and total charge densities at all points on the surface of the sphere. Formul an argument to show thet this clectrie ficld is the actual one.92 ELECTROSTATIC FIELD IN DIELECTRIC MEDIA [cnar. 4 4-14, A uniform electric field Eo is set up in a medium of dielectric constant K. Prove that the field inside a spherical cavity in the medium is 3KEo Er oRTi *4-15. A dielectric sphere of radius R has a permanent polarization P which is uniform in direction and magnitude. The polarized sphere gives rise to an electric field. Determine this ficld both inside and outside the sphere. Inside the sphere the electric field, which is in the opposite direction to the polariza~ tion, is called a depolarizing field. (Hint: Since div P vanishes at all points, the electrostatic potential satisfies Laplace’s equation both inside and outside the sphere. Do not assume that the dielectric is characterized by a dielectric constant.] 4-16, In the text, it was shown that the polarization P = pf( (6+ — 67). Use this relation for the uniformly polarized sphere of Problem 4-15 to de- termine the external dipole field directly.CHAPTER 5 MICROSCOPIC THEORY OF DIELECTRICS In the preceding chapter we were concerned with the macroscopic as- pects of dielectric polarization, and it was shown how in many cases the polarization could be taken into account through the introduction of a dielectric. constant. In this’ way the electric field could be computed directly from a consideration of the free charge distribution. Although reference was made to the molecules of the dielectric several times in Chapter 4, a microscopic treatment of the material was not carried through in detail, and the over-all picture which was presented was certainly from the macroscopic point of view. We should now like to examine the molecu- lar nature of the dielectric, and see how the electric field responsible for polarizing the molecule is related to the macroscopic electric field. Further- more, on the basis of a simple mulecular model it is possible to understand the linear behavior which is characteristic of a large class of dielectric materials. 5-1 Molecular field in a dielectric. The electric field which is responsible for polarizing a molecule of the dielectric is called the molecular field, En. This is the electric field at a molecular position in the dielectric; it is produced by all external sources and by all polarized molecules in the dielectric with the exception of the one molecule at the point under con- sideration. It is evident that E,, need not be the same as the macroscopic electric field because, as was discussed in Section 4-3, the latter quantity is related to the force on a test charge which is large in comparison with molecular dimensions. ‘The molecular field may he calculated in the following way. Let us cut out a small piece of the dielectric, leaving a spherical cavity surrounding the point atewhich the molecular field is to be computed. The dielectric which is left will be treated as a continuum, i.e., from the macroscopic point of view. Now we put the dielectric back into the cavity, molecule by molecule, except for the molecule at the center of the cavity where we wish to compute the molecular field. The molecules which have just been replaced are tg be treated, not as a continuum, but as individual dipoles. The procedure just outlined can be justified only if the result of the calculation is independent of the size of the cavity; we shall see that under certain conditions this is indeed the case. Let us suppose that the thin dielectric sample has been polarized by placing it in the uniform electric field between two parallel plates which are oppositely charged, as shown in Fig. 5-1(a). It-will be assumed that 93,Ron ee ee tot (a) (b) Fic. 5-1. Replacement of the diclectric outside the “cavity” by a system of bound charges. the polarization is uniform on a macroscopic scale (i.e., div P = 0), and that P is parallel to the field producing it. The part of the dielectric external to the cavity may be replaced by a system of bound charges as shown in Fig. 5-1(b), whence the electric field at the center of the cavity vritter may be written as En = Ep + Es +E, +2? 6-1) Here, E, is the primary electric field due to the charged parallel plates, Ez is the depolarizing field due to bound charge on the outside surfaces of the dielectric, E, is due to bound charge on the cavity surface S, and E’ is due to all of the dipoles inside of S. Although we are not concerned with the explicit form of E,, it is evident that if the dimensions of the plates are large compared with their separation, Z, = (1/€9)o, where o is the surface charge density. The depolarizing field is also produced by two parallel planes of charge, this time with the density op. Since op = Py = & P, (6-2) Let us write the macroscopic electric field in the dielectric without a subseript, that is, E. Since the normal component of the electric displace- ment D is continuous across the vacuum-dielectric interface, and since D = € E, in the vacuum just outside the dielectric slab, Ex, = €oE + P. (5-3) Combining Eqs. (5-1), (5-2), and (5-3) yields E, = E+E, +E, (5-4)5-1] MOLECULAR FIELD IN A DIELECTRIC 95 which is an equation relating the molecular field to the macroscopic electric field in the dielectric material. This result is quite general, and not, re- stricted to the geometry of Fig. 5-1; nevertheless, the above derivation is instructive and will be useful to the subject discussed in Section 5~4. The field E, arises from bound charge density, op = Pp, on the spherical surface S.. Using spherical coordinates, and taking the polar direction along the direction of P, as in Fig. 5-2, we obtain ae, P0088) 5 aa, where r is the vector from the surface to the center of the sphere. From symmetry, it is evident that only the component of dE, slong the direction of P will contribute to the integral of (5-5) over the complete surface. Since de == r? sin 69 d¢, . roe ope P| of cos* 8 sin 6d o oO 6-6) Fig, 5-2. Calculation of the “cavity” surface contribution to En. Finally, we come to the last term in (5-4), that due to the electric dipoles inside S. There are a number of important cases for which this term vanishes. If there are a great many dipoles in the cavity, if they are oriented parallel but randomly distributed in position, and if there are no correlations between the positions of the dipoles, then E’ = 0. This is the situation which might prevail in a gas or a liquid. Similarly, if the dipoles in the cavity are located at the regular atomic positions of a eubic erystal,* then again B’ = 0. In this connection, the reader is referred to Problem 5-2. * Crystals with the highest symmetry belong to the cubie system.96 MICROSCOPIC THEORY OF DIELECTRICS [cnar. 5 In the general case, E’ is not zero, and if the material contains several species of molecule, E’ may differ at the various molecular positions. It is this term which gives rise to the anisotropic electrical behavior of calcite, for example. It is not our purpose, however, to develop a theory of aniso- tropic materials; hence we restrict further discussion to the rather large class of materials in which E’ = 0. Thus, Eq. (5-4) reduces to E, = E +36 P. (6-7) It is interesting to note that this result would be obtained directly by the above method if the spherical “cavity” were created by removing just one molecule. But under these conditions the cavity would be so small that the replacement of the rest of the dielectric by a system of bound charges could not be justified. The dipole moment of a molecule per unit polarizing field is called its polarizability, a. In other words, Pm = oEm. (5-8) If there are N molecules per unit volume, then the polarization P = NPm, and combining this result with (5-7) and (5-8), we obtain p= Na(e +P). 9) Beo This equation may be rewritten in terms of the dielectric constant, K, since P = (K — 1)€pE. In this way, Eq. (5-9) becomes’ 83e9 (K — 1) 5o G2, (5-10) which is known as the Clausius-Mossotti equation. It is evident that (5-16) defines a molecular property, namely, the molecular polarizability, in terms of quantities which can be determined on a macroscopic basis. 5-2 Induced dipoles. A simple model. ‘The molecules of a dielectric may be classified as polar or nonpolar. A polar molecule is one which has a permanent dipole moment, even in the absence of a polarizing field En. In the next section the response of a polar dielectric to an external electric field will be studied, but here we deal with the somewhat simpler problem involving nonpolar molecules, in which the “centers of gravity” of the positive and negative charge distributions normally coincide. Sym- metrical molecules such as H», Nz, and Oz, or monatomic molecules such as He, Ne, and A, fall into this category. The application of an electric field causes a relative displacement of the positive and negative charges in nonpolar molecules, and the molecular